READOUT ELECTRONICS FOR MICROBOLOMETER
INFRARED FOCAL PLANE ARRAY
QIAN XINBO
NATIONAL UNIVERSITY OF SINGAPORE
2004
READOUT ELECTRONICS FOR MICROBOLOMETER
INFRARED FOCAL PLANE ARRAY
QIAN XINBO (M. Sc., CAS)
A THESIS SUBMITTED
FOR THE DEGREE OF DOCTOR OF PHILOSOPHY
DEPARTMENT OF ELECTRICAL & COMPUTER ENGINEERING
NATIONAL UNIVERSITY OF SINGAPORE
2004
Acknowledgements_____________________________________________________
Acknowledgements
I wish to express my sincere gratitude to my supervisors, A/Prof. Yong Ping Xu and
A/Prof. Gamani Karunasiri, for their interest, guidance and support throughout the
course of this work. Their profound knowledge and rigorous scientific approach have
greatly inspired me in the research. Their advices to me on all matters, not just the
professional help, are greatly appreciated and are something I will ever benefit from.
My thankfulness also extends to my senior colleagues Mr. H.-J. Wu, Mr. X.-B. He
and Mr. C. Y. Hui. They have greatly helped me in the research. Especially at the
initial stages, I got a lot of support and helpful suggestions through the technical
discussions with them. I would also thank Ms. Musni Hussain, Mr. Francis Boey and
Mr S. M. Teo for their assistance in the experimental set-up. My appreciation also
goes to Dr. V. Samper of Institute of Microelectronics (IME) for providing me the
microbolometer samples and focal plane arrays.
Many thanks to graduate students in Center for Opto-electronics and VLSI Lab, in
particular, Mr. L.-F. Zhou, Ms X.-W. Zhang, Ms. L.-C. Xu and Mr. Rama Krishna
MVS for their kind help. I really enjoy and appreciate the time I have spent with
them.
Finally, I would like to thank my family for their continuously support and
encouragement.
i
Table of Contents_______________________________________________________
Table of Contents
Acknowledgments-------------------------------------------------------------------------------i
Table of Contents-------------------------------------------------------------------------------ii
Summary----------------------------------------------------------------------------------------vii
List of Figures---------------------------------------------------------------------------------ix
List of Tables-----------------------------------------------------------------------------------xv
List of Symbols--------------------------------------------------------------------------------xvi
List of Abbreviations-----------------------------------------------------------------------xviii
Chapter 1. Introduction-----------------------------------------------------------------------1
1.1. Infrared detectors ----------------------------------------------------------------------1
1.2 Microbolometer IR detector and focal plane array --------------------------------2
1.3 Detector characterization--------------------------------------------------------------4
1.4. Structure of micromachined microbolometer --------------------------------------7
1.5. Readout circuits and self-heating effect --------------------------------------------9
1.6. Scope and organisation of thesis----------------------------------------------------11
Chapter 2. Literature Review--------------------------------------------------------------14
2.1. Infrared detectors for thermal imaging---------------------------------------------14
2.2. Microbolometer IR detector---------------------------------------------------------16
2.2.1. Microbolometer IR detector --------------------------------------------------16
2.2.2. Microbolometer Focal plane array-------------------------------------------18
ii
Table of Contents_______________________________________________________
2.3. Readout electronics for microbolometer FPAs -----------------------------------20
2.3.1. Front-end signal conditioning circuits---------------------------------------21
2.3.1.1. Voltage-mode front-end readout circuits------------------------22
2.3.1.2. Current-mode front-end readout circuits-------------------------25
2.3.1.3. Other front-end readout circuits ----------------------------------31
2.3.2. On-chip non-uniformity correction------------------------------------------31
2.3.3. Other on-chip functions -------------------------------------------------------33
2.4. Self-heating effect and cancellation techniques-----------------------------------36
2.5. Summary-------------------------------------------------------------------------------39
Chapter 3. Electro-thermal Characteristics of Microbolometer--------------------40
3.1. Thermal characteristics of microbolometer ---------------------------------------41
3.2. Construction of SPICE electro-thermal model for microbolometer -----------42
3.2.1. Elector-thermal model of microbolometer ---------------------------------43
3.2.2. Verification of the SPICE microbolometer model-------------------------45
3.3. Study of response of the microbolometer to the infrared and self-heating
power-----------------------------------------------------------------------------------48
3.4. Summary ------------------------------------------------------------------------------56
Chapter 4. Readout IC for Microbolometer with Voltage-mode Self-heating
Cancellation---------------------------------------------------------------------57
4.1. A new self-heating cancellation method-------------------------------------------57
4.1.1. Self-heating cancellation circuit----------------------------------------------58
4.1.2. Determination capacitor and current values--------------------------------60
4.1.3. Function verification-----------------------------------------------------------63
iii
Table of Contents_______________________________________________________
4.2. Self-heating cancellation circuit design -------------------------------------------64
4.2.1. Circuit design-------------------------------------------------------------------64
4.2.2. Charge injection and clock feed-through -----------------------------------66
4.2.3. Final simulation results--------------------------------------------------------69
4.3. Incorporation the voltage-mode self-heating cancellation circuit into
ROIC-----------------------------------------------------------------------------------70
4.3.1. Clock and control signals generation---------------------------------------72
4.3.1.1. Clock generator for main_clk and φ1, φ2-------------------------73
4.3.1.2. Clock generator for A/D converter-------------------------------75
4.3.1.3. Pixel addressing signal generation--------------------------------77
4.3.2. Pre-amplifier design-----------------------------------------------------------79
4.3.2. Fixed pattern noise correction -----------------------------------------------83
4.4. Summary-------------------------------------------------------------------------------87
Chapter 5. Readout IC for Microbolometer with Current-mode Self-heating
Cancellation----------------------------------------------------------------------88
5.1. Current-mode self-heating cancellation circuit------------------------------------88
5.1.1. Current-mode self-heating cancellation scheme---------------------------88
5.1.2. Determination of the capacitor and current values ------------------------91
5.1.3. Function verification ----------------------------------------------------------92
5.2. Current-mode ROIC with self-heating cancellation ------------------------------95
5.2.1. Digital circuitry of the current-mode ROIC--------------------------------97
5.2.2. Current subtraction circuit----------------------------------------------------98
5.2.3. Current steering DAC design -----------------------------------------------102
5.2.3.1. Selection of the current-steering DAC structure--------------102
iv
Table of Contents_______________________________________________________
5.2.3.2. Current sources and mirrors--------------------------------------104
5.2.3.3. Decoder logic ------------------------------------------------------109
5.2.3.4. Differential switches and low-crossing latch------------------111
5.2.3.5. Current-steering DAC circuit and performance---------------113
5.2.4. Current-mode fixed pattern noise cancellation circuit-------------------117
5.3. Summary ------------------------------------------------------------------------------121
Chapter 6. Experimental Results--------------------------------------------------------122
6.1. Microbolometers used in measurement ------------------------------------------123
6.2. Test of the voltage-mode self-heating cancellation circuits with single
microbolometer----------------------------------------------------------------------125
6.3. Test of the current-mode self-heating cancellation circuits with single
microbolometer----------------------------------------------------------------------129
6.4. Test of the voltage-mode ROIC with an external microbolometer FPA-----132
6.4.1. Test set-up of the voltage-mode ROIC ------------------------------------132
6.3.1.1. Hardware interface of the test system --------------------------134
6.3.1.2. Image acquisition software---------------------------------------138
6.3.2. Test results of the voltage-mode ROIC------------------------------------140
6.4. Test of the current-mode fixed pattern correction circuit on ROIC ----------147
6.5. Summary -----------------------------------------------------------------------------151
Chapter 7. Conclusion----------------------------------------------------------------------153
7.1. Conclusions----------------------------------------------------------------------153
7.2. Original contributions----------------------------------------------------------154
7.3. Future work----------------------------------------------------------------------154
v
Table of Contents_______________________________________________________
References-------------------------------------------------------------------------------------156
Appendix A: List of Publications---------------------------------------------------------174
Appendix B: LabVIEW programs-------------------------------------------------------175
B.1. Descriptions of the sub-VIs used in imaging acquisition programs-----175
B.2. LabVIEW program for calibration operation-------------------------------176
B.3. LabVIEW program for real-time imaging----------------------------------179
B.4. LabVIEW program for noise measurements-------------------------------182
vi
Summary_____________________________________________________________
Summary
Uncooled microbolometer infrared detectors have a number of advantages over other
types of detectors. They have showed their commercial potential to realize low-cost,
high-performance infrared imagery systems. Microbolometer changes its resistance in
response to the infrared radiation. In order to monitor the change of the resistance,
microbolometers need to be electrically biased during the operation. Unfortunately,
the bias current flowing through the microbolometer generates the heat on the sensor
itself. This unavoidable self-heating effect will severely affect the readout circuits,
such as the degradation of the dynamic range. The continuing improvement on the
performance of the uncooled thermal imagers have driven the motive of developing
an effective circuit for self-heating cancellation that can be applied to on-chip readout
electronics for the microbolometer focal plane array.
This project looks into a novel technique of self-heating cancellation for
microbolometer. Such a cancellation scheme is based on the equivalence between the
electrical and thermal systems, and uses a capacitor to mimic the thermal capacitance
of the microbolometer. A replica of the bias heating can generated on the capacitor
and later used to cancel the self-heating of the bolometer. The concept of using the
electrical equivalence of bolometer’s thermal parameter to build the replica of self-
heating is extended to current-mode readout circuits. A PMOS transistor working in
the triode region is used to generate the ramp current to compensate the self-heating
effect in a current subtraction circuit. Both voltage- and current-mode self-heating
cancellation circuits are realized in silicon and the effectiveness is demonstrated with
a single microbolometer.
vii
Summary_____________________________________________________________
The voltage- and current-mode ROICs (readout integrated circuits) with the proposed
self-heating cancellation circuit have been developed together with on-chip 8-bit fixed
pattern noise (FPN) correction circuits. The ROICs were fabricated in a 0.6 µm
CMOS technology. The voltage-mode ROIC was evaluated with an external 128 ×
128 microbolometer FPA. The test results have shown that the FPN can be corrected
to a 7-bit resolution when exclude the pixels with unexpected large FPN from test,
and an 8-bit resolution can be attained when tested with a built-in resistor array. The
current-mode ROIC has been evaluated with a built-in 8 × 8 resistor array for the
performance of FPN correction, since the microbolometer FPA for current-mode
ROIC is not available. The measured residual FPN of less than ±2.5 LSB has been
achieved. Due to the unavailability of proper microbolomter FPAs, the self-heating
cancellation cannot be demonstrated with the on-chip FPN correction in the ROICs.
viii
List of Figures ________________________________________________________
List of Figures
Fig. 1.1. Block diagram of a bolometer infrared detection system------------------------4
Fig. 1.2. Thermal image taken by the uncooled infrared camera LTC550---------------4
Fig. 1.3. Structure of a bulk micromachined bolometer------------------------------------8
Fig.1.4. Microstructure of the surface micomachined microbolometer-------------------9
Fig.1.5. Basic readout circuit of the microbolometer--------------------------------------10
Fig.1.6. Wheatstone bridge readout circuit of the microbolometer----------------------10
Fig. 2.1. Flip-chip-hybrid (a) and monolithic (b) structures of microbolometer FPA ------------------------------------------------------------------------------------19
Fig. 2.2. Block diagram of the ROIC for an M × N-pixel microbolometer focal plane array----------------------------------------------------------------------------21
Fig. 2.3. Two simplest voltage-mode readout circuits-------------------------------------23
Fig. 2.4. Chopper-amplifier voltage-mode circuit------------------------------------------23
Fig. 2.5. AC-bias bridge readout circuit for bolometer------------------------------------24
Fig. 2.6. Bridge readout circuit with capacitive load --------------------------------------25
Fig. 2.7. A simple current-mode readout for microbolometer----------------------------26
Fig. 2.8. Pixel amplifier and pulse current bias circuits-----------------------------------27
Fig. 2.9. Direct injection circuit---------------------------------------------------------------28
Fig. 2.10. Direct injection circuit with CTIA integrator-----------------------------------28
Fig. 2.11. The shared integration cell with current-mode background suppression---29
Fig. 2.12. CCBDI readout circuit-------------------------------------------------------------30
Fig. 2.13. Pixel-wised A/D converter circuit-----------------------------------------------31
Fig. 2.14. Block diagram of the two-mode 16-bit A/D converter------------------------33
Fig. 2.15. Method for correction for temperature-induced non-uniformities with thermally shorted microbolometer elements -----------------------------------34
ix
List of Figures ________________________________________________________
Fig. 2.16. Block diagram of the three-point correction scheme--------------------------35
Fig. 2.17. Front-end readout circuit with the self-heating cancellation using blind bolometer----------------------------------------------------------------------------37
Fig. 2.18. Self-heating compensation circuits in [131] [132]-----------------------------37
Fig. 2.19. Off-chip self-heating compensation schemes: (a) Compensation of the difference of Vref and Vd; (b) Direct compensation of Vd.---------------------39
Fig. 3.1. Electro-thermal model of microbolometer ---------------------------------------44
Fig. 3.2. Experimental setup in Ref.[95], where RB is the microbolometer and R0s are the fixed reference resistors.----------------------------------------------------45
Fig.3.3. Comparison of simulation results using the SPICE model with the experimental data ---------------------------------------------------------------------47
Fig. 3.4. Wheatstone bridge readout circuit-------------------------------------------------50
Fig. 3.5. Temperature changes of the microbolometer, ∆T, with different IR powers---------------------------------------------------------------------------------51
Fig.3.6. Output voltage (green curves) and temperature change (pink curves) of the the microbolometer with different input IR powers. ----------------------------52
Fig. 3.7. Output voltage (green curves) and temperature change (pink curves) of the microbolometer with different input IR powers, at constant current bias.-----55
Fig. 3.8. Output current (green curves) and temperature change (pink curves) of the microbolometer with different input IR powers, at constant voltage bias -----55
Fig. 4.1. Schematics of the self-heating canceling scheme -------------------------------59
Fig. 4.2. Waveforms of the clock signals----------------------------------------------------59
Fig. 4.3. (a) Half circuit of the Wheatstone bridge containing the microbolometer and (b) I-C pair that generates the replica of the self-heating------------------60
Fig. 4.4. Mimic the self-heating effect using an I-C pair----------------------------------62
Fig. 4.5. Simulation results for the self-heating cancellation circuit---------------------64
Fig. 4.6. The schematics of the tuneable self-heating cancellation circuit--------------65
Fig. 4.7. Output of the implemented self-heating cancellation circuit with non- ideal switches-------------------------------------------------------------------------66
Fig.4.8. Charge injection vs. different aspect ratio of M4. The upper plot shows non- inverting inputs of the preamplifier, while the lower plot shows differential inputs of the preamplifier, at the different aspect ratio of M4-------------------68
x
List of Figures ________________________________________________________
Fig.4.9. Charge injection after the size of M4 is optimized. The upper plot shows non- inverting inputs of the preamplifier, while the lower plot shows differential inputs of the preamplifier, when size of M4 over M3,5 are around 2.3--------68
Fig.4.10. Simulation results of optimized self-heating cancellation circuit-------------70
Fig. 4.11. Partially parallel readout architecture of the voltage-mode ROIC-----------71
Fig. 4.12. Block diagram of the fixed pattern noise correction circuit ------------------72
Fig. 4.13. Block diagram of clock and control signal generator--------------------------73
Fig.4.14. Schematic of the pixel clock and control signal generator --------------------74
Fig.4.15. Waveforms of int_ctrl, φ2, φ1 main_clk and clk -------------------------------75
Fig.4.16. Schematic of the ad_clk and ad_start signal generator------------------------76
Fig.4.17. Timing waveforms of ad_start, ad_clk, frame, int_ctrl and main_clk-------76
Fig. 4.18. Block diagram of the X-Y scanning clock generator--------------------------78
Fig. 4.19. Waveforms of the scanning clocks of the first 6 rows and main_clk -------78
Fig. 4.20. Waveforms of the scanning clocks of the first 6 columns and col_clk -----79
Fig. 4.21. Schematic of the pre-amplifier ---------------------------------------------------82
Fig. 4.22. Layout of the preamplifier --------------------------------------------------------82
Fig. 4.23. Schematic of the FPN correction circuit ----------------------------------------84
Fig. 4.24. Simulation results of fixed pattern noise correction---------------------------87
Fig. 5.1. Schematic of the current-mode self-heating cancellation circuit--------------90
Fig. 5.2. Schematic of the ramp current generator-----------------------------------------90
Fig. 5.3. Simulation results for current-mode self-heating cancellation----------------94
Fig. 5.4. Tuneable feature of the ramp current generator---------------------------------94
Fig. 5.5. Pixel structure of the voltage-mode(a) and current-mode(b) readout circuit----------------------------------------------------------------------------------95
Fig. 5.6. Readout circuit architecture of the current-mode ROIC for an M × N FPA------------------------------------------------------------------------------------96
Fig. 5.7. X-Y address clock generator for M × N FPA -----------------------------------97
xi
List of Figures ________________________________________________________
Fig. 5.8. Current subtraction circuit ---------------------------------------------------------98
Fig. 5.9. (a) AC sweep response and (b) DC sweep of the current subtraction circuit-------------------------------------------------------------------------------100
Fig.5.10. Noise analysis of the current subtraction circuit: (a) Equivalent input current noise density, (b) output noise power density and (c) total noise power in the bandwidth from 1Hz to 1MHz------------------------------------100
Fig.5.11.(a) Equivalent input current noise density; (b) output noise power density and (c) total noise power in the bandwidth from 1 Hz to 1 MHz------------101
Fig. 5.12. Binary weighted architecture of the current-steering DAC -----------------103
Fig. 5.13. Thermometer coded architecture of the current-steering DAC-------------103
Fig. 5.14. Segmented architecture of the 8-bit current-steering DAC -----------------104
Fig. 5.15. Schematic of the bandgap voltage reference circuit-------------------------107
Fig. 5.16. Temperature dependence of the bandgap reference voltage output--------107
Fig. 5.17. Schematic of the voltage-to-current converter--------------------------------109
Fig. 5.18. Simulation result of the 1 LSB basic current generator----------------------109
Fig. 5.19. 2-bit binary-to-thermometer decoder circuit and latches--------------------110
Fig. 5.20. Waveforms of the decoder inputs, S0, S1 and outputs Line(0-3).----------111
Fig. 5.21. Schematic of the low-crossing latch--------------------------------------------111
Fig. 5.22. Transient analysis results of the low-crossing latch--------------------------112
Fig. 5.23. PMOS differential switches of the current outputs---------------------------113
Fig. 5.24. Block diagram of the current-steering DAC----------------------------------114
Fig. 5.25. Layout of the current-steering DAC--------------------------------------------115
Fig. 5.26. Transient analysis of the current-steering DAC------------------------------116
Fig. 5.27. Simulated DNL of the current-steering DAC---------------------------------116
Fig. 5.28. Simulated INL of the current-steering DAC----------------------------------117
Fig. 5.29. Schematic of the current-mode FPN correction circuit----------------------118
Fig. 5.30. The residual FPN current after current-mode FPN correction--------------121
xii
List of Figures ________________________________________________________
Fig. 6.1. The microphoto of the ROIC-----------------------------------------------------123
Fig. 6.2. (a) SEM photo of the microbolometer on FPA; (b) The microphotograph of the microbolometer FPA chip; (c) On-FPA readout bridge and the output of FPA to our ROIC -----------------------------------------------------125
Fig. 6.3. The test set-up of voltage-mode self-heating cancellation circuit for single bolometer--------------------------------------------------------------------126
Fig. 6.4. Measured output from preamplifier with different I0/C0 ratios---------------128
Fig. 6.5. Measured output from integrator with and without self-heating cancellation.- ------------------------------------------------------------------------------------------128 Fig. 6.6. Measured residue error after self-heating cancellation------------------------128
Fig. 6.7. The test set-up of current-mode self-heat cancellation for single bolometer-------------------------------------------------------------------------------------------128
Fig. 6.8. Measured Id and Iref - Iramp---------------------------------------------------------130
Fig. 6.9. Integrator outputs with and without self-heating cancellation, at input IR power of 175nW. -------------------------------------------------------------------131
Fig. 6.10. Measured residue error (Idiff) after self-heating cancellation----------------131
Fig. 6.11. Tunable feature of the current-mode self-heating cancellation circuit-----131
Fig. 6.12. Test set-up for voltage-mode readout circuit----------------------------------133
Fig. 6.13. The signal flow diagram of the hardware interface of the test system-----135
Fig. 6.14. Schematic of the digital circuit on the PCB-----------------------------------137
Fig. 6.15. Relationship of main_clk, int_clk and Convert*------------------------------138
Fig. 6.16. User interface of image acquisition---------------------------------------------139
Fig. 6.17. Flow chart of the Image Acquisition LabVIEW programs: (a) Real time imaging operation; (b) Calibration operation.--------------------------------140
Fig.6.18. Output waveform chart of the first row of FPA without FPN correction--142
Fig. 6.19. Measured 2-D Intensity graph of the FPA without coarse FPN correction-------------------------------------------------------------------------------------------142
Fig. 6.20. The residual FPN readings of the first raw of FPA---------------------------143
Fig. 6.21. The residual fixed pattern of the FPA after on-chip correction: (a) Intensity scale is same as that in Fig .6.19 and (b) Intensity scale is reduced-----------------------------------------------------------------------------143
xiii
List of Figures ________________________________________________________
Fig. 6.22. The histogram of the residual FPN---------------------------------------------144
Fig. 6.23. Residual error of FPN correction when tested with a resistor array-------145
Fig. 6.24. Test set-up for current-mode FPN correction circuit on ROIC-------------148
Fig. 6.25. Measured residual error after coarse FPN correction------------------------149
Fig. 6.26. Output waveform of the on-chip current integrator. (a) before and (b) after coarse FPN correction--------------------------------------------------150
xiv
List of Tables__________________________________________________________
List of Tables
Table 3.1. Electrical equivalents of the thermal behaviour quantities of microbolometer---------------------------------------------------------------------42
Table 3.2. Parameters of the two bolometers used in the experiments and simulation---------------------------------------------------------------------------47
Table 3.3. Parameters of the microbolometer used in the simulation-------------------50
Table 4.1. The circuit parameters after optimization --------------------------------------69
Table 4.2. Post-layout Simulated pre-amplifier parameters ------------------------------83
Table 5.1. Current-mode self-heating cancellation circuit parameters in simulation---------------------------------------------------------------------------93
Table 5.2. Truth table of the 2-bit binary-to-thermometer decoder--------------------110
Table 5.3. Simulation results of the on-chip current-mode FPN correction circuit-------------------------------------------------------------------------------120
xv
List of Symbols_________________________________________________________
List of Symbols
Symbol Description Unit Page No.
α Temperature coefficient of resistance 1/K 6,43,50,53,60 ~ 63, 91,92
ε Absorptivity coefficient 41
φe Spectral incident infrared power W 5
ϕ Input radiation power W 41
η Emissivity of the detector material 6
pµ Hole mobility cm2/V.s 92
τ Thermal time constant s 41,46~48,58
ω Angular modulation frequency of incident radiation rad/s 5, 6
∆f Noise equivalent bandwidth Hz 6
∆T Temperature difference of the detector K 48,50~53, 55, 60 ~ 63, 91
∆TbTemperature difference between the target and the background K 5
vℜ Responsivity V/W 5, 6
A Sensitive area of the detector cm2 6
B Bandwidth Hz 54
oxC Gate oxide capacitance per unit area f/ cm2 92
CT Electrical equivalent of thermal capacitance f 42,44,46,50
D* Specific Detectivity cm√Hz/W 6
G Thermal conductance W/K 6,42,41,46,48
H Thermal capacitance J/K 6,41,42,48, 61,62,91,92
xvi
List of Symbols_________________________________________________________
Symbol Description Unit Page No.
IS Saturation current of the bipolar device A 105
k Boltzmann’s constant J/K 54,106
Pbias Bias heating power W 41~45,48,60~62,91
PIR Infrared power absorbed by detector W 41~45,52,55,60
q Electron charge C 106
RT Electrical equivalent of Thermal resistance Ω 42,44,46,50
T Temperature K 41,42,54,105,106
VG0 Bandgap voltage of the silicon at 0 ºK V 105,106
VGS, VDS
Gate-source and drain-source voltage of transistor, respectively V 92,98
VT Thermal voltage V 105,106
VTH Threshold voltage V 92
W, L Width and length of the transistor, respectively µm 67,92
xvii
List of Abbreviations____________________________________________________
List of Abbreviations
Abbreviation Expansion Page No.
ADC, or A/D Analog to digital converter 31~33,72~74,77,85,87,96,97,120,133,136,145,146,148
AMS AustriaMicroSystems 72,92
CCBDI Constant current buffered direct injection 30
CMOS Complementary Metal-Oxide Semiconductor 7,8,18,28,63,64,92, 105,122,123,151, 153
CTIA Capacitive transimpedance amplifier 28,29
DAC, or D/A Digital to analog converter 32,34,72,73,85~88,102~105,107,113~117,119,121,133, 148,149
DAQ Data acquisition 132~136,141,146
DI Direct injection 27~30
DNL Differential Nonlinearity 102,116,117
FPA Focal plane array
3,9,11,12,14,17~21, 23,25,30~35,48,56,57,70~73,76~79,85~ 88,95~99,118,119, 122~125,132~136, 140~147,151~155
FPN Fixed pattern noise 31,32,36,57,66,72,79, 83~88,96~99,114, 117~122,133~136, 139,141~155
GBW Gain bandwidth 79,80,83
IC Integrated circuit 14,35,58,63,70,88
IME Institute of Microelectronics 132
INL Integral Nonlinearity 102,117
IR Infrared
1~5,9,12,14~20,28,31~39,48~57,60~63,77, 85,88,91,93~96,118, 119,122,129~134,140,146,147,152~155
JFET Junction field effect transistor 25
xviii
List of Abbreviations____________________________________________________
Abbreviation Expansion Page No.
LSB Least significant bit 104,107~109,113,119,121,145,146,149,151, 154
NEP Noise Equivalent Power 5,7
NETD Noise Equivalent Temperature Difference 6,17,19,20,85
NMOS N-channel Metal-Oxide Semiconductor transistor 64,67,71,95,107
PCB Printed circuit board 129,132,134~138,147
PMOS P-channel Metal-Oxide Semiconductor transistor 22,64,80,90,92,105,109~113,121
PTAT Proportional to absolute temperature 106
RMS Root mean square 5
ROIC Readout integrated circuits
9,12,14,19~22,31 ~34,39,70,71,87,95~ 98,122~125,129, 132~136,140~142, 144,146~148,151~155
SEM Scanning electron microscope 125
SNR, or S/N Signal-to-noise ratio 5,6,54,124,132,147
SPICE Simulation program with integrated circuit emphasis
12,42,44~49,56,62, 92,153
SRAM Static Random Access Memory 135,138,141,147,148
TCR Temperature coefficient of resistance 6,7,17,43
VCVS Voltage control voltage source 63
VLSI Very large scale integration 62,63
xix
Chapter 1.__________________________________________________Introduction
Chapter 1. Introduction
1.1. Infrared detector
Infrared (IR), or “beyond the red”, is the electromagnetic radiation similar to visible
light but with longer wavelengths. The infrared portion of the electromagnetic
spectrum lies adjacent to the visible[1]. Its wavelength range extents from 0.7µm, the
red end of the visible light, to about 1000µm. While visible light is only emitted by
objects at a very high temperature, infrared energy is radiated by all objects when the
temperature of the object surface is above absolute zero. For example, at room
temperature, the peak of the radiant emittance is about 10 µm which lies in the
infrared region[2][3]. Hotter the object, more energy is radiated.
Although infrared radiation is invisible to the naked human eyes, it can be “seen” by
infrared detectors. It gives us the possibility of measuring the self-emitted radiant
energy from objects. When the infrared radiation is effectively detected, we can see
all objects regardless of the darkness of the environment.
An infrared detector is basically a radiation transducer. It senses the incident infrared
radiation with the changes in some of its electrical parameters. The infrared detectors
can be roughly classified into two groups: photon and thermal detectors. The former
absorbs the incident photons, which interact with the charge carriers in the detector
and generates a photo-current or voltage, which produce the electrical output directly
proportional to the number of input photons. The thermal detector responds to the
1
Chapter 1.__________________________________________________Introduction
infrared power with a change in its temperature, which can be converted to either a
voltage or current signal through a temperature sensing device.
The photon detector needs to be cooled down with a cryogenic system for efficient
operation. However, most of thermal detectors are “uncooled”, which means that they
can operate at room temperature. This enables uncooled detectors to have some
advantages over the photon ones, such as affordability and reliability. These attractive
characteristics encourage the development of the uncooled IR for a wide range of
applications.
Since William Herschel first discovered the infrared energy beyond the visible region,
the use of this portion of electromagnetic spectrum has aroused great interest among
scientists. Modern infrared imaging system are lighter in weight and easier to use
than those used decades ago[4][5]. They can sense the infrared radiant energy and
generate useful electrical signal that is proportional to the temperature of the object
surface. Therefore, clear thermal images can be produced for visual inspection.
Nowadays, infrared technology is being widely adopted in many military, civilian,
and scientific applications[6~8].
1.2. Microbolometer IR detector and focal plane array
There have been several types of uncooled IR detectors reported to date. Among them,
microbolometer IR detector has been widely adopted in commercial products owing
to its low cost and compatible process with silicon integrated circuits.
2
Chapter 1.__________________________________________________Introduction
Microbolometer is a temperature sensor whose resistance is dependent on its
temperature. A microbolometer based infrared detector consists of a resistive
elements built on a thermally isolated structure, which is also act as infrared
absorption layer. The incident infrared radiation is absorbed by the detector, causing
its temperature to rise. The temperature change of the detector is sensed by the change
in resistance. With the readout circuit, the resistance change is converted to a voltage
or current signal as the output of the detector.
When used in the thermal imaging applications, microbolometer detectors are
commonly arranged in a two-dimensional detector array, which is called the focal
plane array (FPA). A FPA is a detector array placed at the focal plane of an optical
system such as a camera, spectrometer or telescope. This detector array is combined
with readout circuits or multiplexer which allows the electronic access to each
detector cell in the array. Each detector in FPA represents a pixel of the final thermal
image.
The operation of a microbolometer imaging system starts from the infrared incident
absorption by the microbolometer FPA. Each detector cell senses the infrared energy
radiated on it and changes its resistance accordingly. The readout electronics of the
IR FPA sequentially addresses the detector cells (pixels) and convert the resistance
change to an electrical signal, either in a voltage or a current. The signal from
detectors is processed by the pre-amplifier and then analog-to-digital converter. After
some imaging process programs such as filtering and enhancement, the digitized
thermal image information can be obtained and stored in the video memory. Thus, a
complete thermal image with the proper format and frame rate for TV display can be
monitored.
3
Chapter 1.__________________________________________________Introduction
The block diagram of a microbolometer IR imaging system is illustrated in the Fig.1.1.
An example of a thermal image taken by the microbolometer based infrared camera is
shown in the Fig.1.2, where a bright spot corresponds to the high temperature and a
dark spot corresponds to the low temperature.
Fig.1.1. Block diagram of a bolometer infrared detection system.
Fig.1.2. Thermal image taken by the uncooled infrared camera LTC550[9].
1.3. Detector characterization
4
Chapter 1.__________________________________________________Introduction
In the study of thermal detector based systems, the following parameters are
commonly used for characterizing the detectors.
• Responsivity:
The basic function of a detector is to convert input IR radiant power to an
electrical output signal. The responsivity, as a measure of the conversion efficiency,
is defined as the output signal per unit input IR power.
e
sv
Vφ
=ℜ (1-1)
Where Vs is the detector output voltage and φe is the spectral incident power, which
can normally be decomposed into a set of sinusoidal components as
tjee e ωφφ 0,= (1-2)
Ideally, the responsivity of the microbolometer detector is broad and flat with respect
to the wavelength of input radiation.
• Noise Equivalent Power (NEP):
NEP is defined as the root mean square (RMS) incident radiant power required to
produce an output signal Vs that is equal to the detector noise level Vn.
ns
e
VVNEP
/φ
= . (1-3)
In other words, the NEP determines the incident power for the detector to produce a
signal-to-noise ratio (SNR) of unity. It is a measure of the ultimate sensitivity of a
detector.
5
Chapter 1.__________________________________________________Introduction
• Specific Detectivity (D*):
The specific detectivity, D*, which is the sensitivity normalized to the area of
1cm2 and the noise equivalent bandwidth of 1Hz, is defined as
NEPfA
D∆
=* , (1-4)
where A is the sensitive area of the detector and ∆f is the noise equivalent bandwidth.
• Noise Equivalent Temperature Difference (NETD):
NETD is defined as the smallest temperature difference between target and the
background, which produces a signal to noise ratio (SNR) of unity.
ns
b
VVT
NETD/
∆= (1-5)
∆Tb is the temperature difference between the target and the background.
When a bridge readout circuit is used, the voltage responsivity of microbolometer can
be expressed as
( ) 2/12224 HG
Vbiasv
ω
αη
+
⋅=ℜ (1-6)
where α is the temperature coefficient of resistance (TCR), η is the emissivity of the
detector material, G is the thermal conductance, H is the thermal capacitance of the
microbolometer, ω is angular frequency of incident radiation. The TCR is defined as
d
d
d dTdR
R1
=α (1-7)
where Rd and Td are the resistance and temperature of the detector, respectively.
6
Chapter 1.__________________________________________________Introduction
It can be seen from Eq.(1-6) that, to achieve high responsivity, a microbolometer with
high TCR, high absorption coefficient and low thermal conductance is desirable. This
is also true for obtaining low noise equivalent power (NEP) and high detectivity (D*).
Moreover, the responsivity of the bolometer has a low-pass feature. Although low
thermal conductance can increase the responsivity, it reduces the bandwidth of the
detector, and hence slows down the response of the microbolometer. Thus,
compromise needs to be made in the detector design.
1.4. Structure of micromachined microbolometer
According to the Eq.(1-6) and Eq.(1-7), the thermal isolation between the
microbolometer and the substrate is one of the important factors that contributes to
high performance of the detector. With the development of silicon micromachining
technology, structure with good thermal isolation can be easily realized. The
compatibility of silicon micromachining technology with the standard CMOS process
offers the possibility of combining the conventional CMOS readout integrated circuits
with the microbolometer focal plane array.
There are two main types of micromachining (bulk and surface) commonly used for
fabricating microbolometers. The structure of a microbolometer made by bulk
micromachining is schematically shown in Fig. 1.3. An infrared absorber layer is
formed by the dielectric layers, normally the composite of silicon oxide and silicon
nitride. A meandering thin film resistor material is then deposited on the dielectric
layer. The silicon under the dielectric membrane is etched away later in the process to
form the gap between the microbolometer and substrate in order to reduce the thermal
7
Chapter 1.__________________________________________________Introduction
mass and the thermal conductance. The resulting micromembrane is suspended above
the silicon substrate and supported by the two legs connecting the detector to the
readout circuits. The absorbed heat can only be dissipated through these two legs to
the substrate that acts as a heat-sink.
Fig. 1.3. Structure of a bulk micromachined bolometer.
In surface micromachining technology, a sacrificial silicon dioxide layer is deposited
at the early stage of the processing. Then the microbolometer thin film is fabricated
on the top of the sacrificial layer. The infrared absorbing layer, similar to that of the
bulk micromachined structure, is made of Si3N4 during the CMOS process. Finally,
the oxide sacrificial layer is removed with either wet or dry etching techniques. The
free-standing microbolometer structure is then formed, as shown in Fig.1.4. After the
sacrificial layer is removed, an air gap of 0.5 µm separates the micromembrane from
the silicon substrate. The absorbed heat is dissipated to the substrate through the
anchoring contacts.
8
Chapter 1.__________________________________________________Introduction
Fig.1.4. Microstructure of the surface micomachined microbolometer.
1.5. Readout circuits and self-heating effect
Besides the microbolometer detector itself, the readout integrated circuit (ROIC) is
the next crucial part of the IR imaging system. It is an interface between the IR
detector array and the thermal image display, where the signal and imaging process
are carried out. The basic configuration of the readout circuit for microbolometer
FPA is shown in Fig.1.5, where Rd is the microbolometer which is biased with a
constant voltage Vbias through a load resistor, R0. The IR radiation is sensed by Rd
and a small voltage change ∆Vd is produced, and subsequently amplified by the
preamplifier. The main shortcoming of this configuration is the large pedestal voltage
Vd is also amplified such that it overrides the IR signal.
9
Chapter 1.__________________________________________________Introduction
Fig.1.5. Basic readout circuit of the microbolometer.
Fig.1.6. Wheatstone bridge readout circuit of the microbolometer.
10
Chapter 1.__________________________________________________Introduction
Most practical readout circuits employ the Wheatstone bridge structure[10], as shown
in Fig.1.6. When the ratio of the resistor R0 over the room temperature resistance of
the microbolometer, Rd0, is equal to that of resistor R1 over R2, the bridge is balanced.
Such a bridge structure not only removes the pedestal bias voltage, but also rejects the
fluctuation and other noise introduced by the power supply.
However, the Wheatstone readout circuit also has its shortcomings, such as self-
heating effect. During the readout, in order to monitor the amount of the resistance
change, the microbolometer element must be electrically biased. When there is a
current flowing through the bolometer, it generates the Joule heating. As a result,
the microbolometer is not only heated by the infrared radiant energy, but also the bias
current flowing through. This heating effect due to the bias current is referred to as the
self-heating of the microbolometer. Because of the micromachined structure, thermal
conductance of the bolometer is greatly reduced. Hence the heat generated by the
self-heating effect cannot be quickly dissipated through the thermal conduction to the
substrate. This makes the temperature change of the microbolometer due to the self-
heating effect much greater than that due to the input infrared radiation. Thus, the
self-heating effect must be compensated in the readout circuits.
1.6. Scope and organization of thesis
Several microbolometer based IR focal plane detector arrays have been reported.
However, very little information is available on their readout electronics, especially
on self-heating cancellation. An in-house microbolometer focal plane array was
developed by previous students. However, its performance suffers from the problems
11
Chapter 1.__________________________________________________Introduction
in its ROIC, in particular, the self-heating effect, as it was not dealt with in the ROIC.
The scope of this project is to develop a simple and yet effective method or circuit for
the self-heating cancellation and employ it in the ROIC for the microbolometer focal
plane array made in house. The overall objective was to develop a prototype ROIC
for the above mentioned microbolometer focal plane array that can compensate both
self-heating and fixed pattern noise.
The organization of the thesis is as follows:
Chapter 2 presents the literature review of the infrared imaging system and readout
electronics for microbolometer focal plane array.
In chapter 3, a SPICE Electro-thermal model of the bolometer is introduced. The
effectiveness of this model is demonstrated by the comparison between the simulation
and the experimental results. The analysis of the thermal behaviours of the
microbolometer is studied based on the proposed SPICE model.
In chapter 4, the design of the voltage-mode readout integrated circuits for
microbolometer FPA is presented. A new self-heating cancellation scheme is
proposed and the detailed description of the circuit operation is given. The proposed
self-heating cancellation scheme is employed in a voltage-mode ROIC, in which the
fixed pattern noise cancellation circuit is also included. Finally, the simulation results
are presented.
Chapter 5 focuses on the design of the current-mode readout electronics for the
microbolometer IR focal plane array readout circuit. The self-heating cancellation
12
Chapter 1.__________________________________________________Introduction
scheme described in chapter 4 is extended to the current-mode realization. The design
and simulation results are presented.
Chapter 6 deals with test and evaluation of the voltage- and current-mode readout
circuits. Finally, chapter 7 gives the conclusions of this research and suggestions for
the future work.
13
Chapter 2.______________________________________________Literature Review
Chapter 2. Literature Review
Over a century after the infrared is discovered, the first IR imaging system circa-1930
Evaporagraph [11] was produced. It was a rather insensitive non-scanning system
which could not satisfied most tasks. Since then, rapid advances have been made in
thermal imaging systems. Today, the sophisticated thermal imaging systems are used
in many commercial, industrial and military applications [12][13]. The ever-
expanding field of applications of IR imaging systems is a primary motivation for the
worldwide efforts in the research and development of infrared technology.
Infrared detectors and readout integrated circuits (ROIC) are the two important
aspects of thermal imaging systems. In late 1970s, the first generation IR focal plane
arrays incorporating the IC technology were developed [14][15]. The readout
integrated circuits for these IR FPA were simple multiplexers and consist of little, if
any, signal conditioning circuits. Due to the significant system performance
enhancements allowed by the integrated circuits, ROICs for IR imaging systems
become more and more complex. The available on-ROIC or on-FPA functionality
has increased dramatically. The advance in modern microelectronic technologies has
provided the potential to realize cost-effective and high performance IR imaging
systems.
2.1. Infrared detectors for thermal imaging
14
Chapter 2.______________________________________________Literature Review
IR detectors are generally classified into two categories, namely, photon and thermal
IR detectors. Photon detectors absorb the IR radiation by interaction with electrons
within the material. When an electron gains enough energy from the incident IR
radiation (or photons), it overcomes the energy gap and becomes a free electron that
alters the electrical parameters of the material, such as conductivity. Thus, the
responsivity of the photon detector depends on the wavelength of the IR radiation, as
the incident photon energy is a function of the wavelength. For a detector material
with a fixed energy gap, when the wavelength of the incident IR radiation is beyond
certain value, the photon energy will not be high enough to excite the electrons to
over the energy band gap. The detector will fail to response. This is often referred to
as “long wavelength limit” of the IR photon detectors. The major advantages of the
photon detector are its good sensitivity and fast response[16][17]. However, in order
to achieve this, the cryogenic cooling operation is required, which makes the
semiconductor photon detectors bulky, heavy and inconvenient to use, as well as
costly. Thus the applications of the photon IR detectors are mainly in the high-
performance thermal imaging system or cameras[18~23].
Thermal IR detectors, on the other hand, absorb the IR radiant power resulting in the
change of its temperature. This temperature change is converted to an electrical output
through a temperature sensing device. Thermal effects are generally wavelength
independent. The spectral band of thermal detectors is normally determined by optics
and windows. Thus, a broad or flat spectral response can be expected. Since the room
temperature object emits the highest radiant power in 8-14µm range, thermal detectors
are generally used as long wavelength IR detector.
15
Chapter 2.______________________________________________Literature Review
In comparison with photon detectors, thermal detectors are characterized by modest
sensitivity and slow response. However, the thermal detectors are typically operated
at the room temperature, making them low-cost and convenient to use. The thermal IR
detectors are often referred to as uncooled IR detectors. Uncooled IR detector
technology has advanced rapidly in the past decade and results dramatic improvement
in the performance of uncooled thermal imaging systems [24][25]. The crucial
technology to fabricate uncooled detectors is the micromachining technology. It
enables the realization of miniaturized thermal isolated structure and increases the
detector sensitivity and response speed[26~29]. The most common uncooled thermal
detectors are microbolometer[30~33] and ferroelectric IR detectors[34~39].
As the microbolometer IR detector will be used in this study, the remaining of this
chapter briefly reviews microbolometer IR detectors and its readout electronics.
2.2. Microbolometer
Microbolometer detector is one of the promising thermal IR detectors. When exposed
to IR energy, the microbolometer detector heats up and its electrical resistance
changes in response to the absorbed IR energy. The resistance change can be
measured by applying an electrical bias to the detector and readout the changes in
voltage or current.
2.2.1. Microbolometer IR detectors
One of the successful commercialized microbolometer materials is vanadium
oxide[40~42], VOx, which was developed originally at Honeywell. It usually has the
16
Chapter 2.______________________________________________Literature Review
high temperature coefficient of resistance (TCR) of above - 2%/K[43]. The measured
thermal capacitance of VO2 detectors is about 10-9 J/K corresponding to a thermal
time constant of 10 ms. The responsivity of 250 kV/W has been observed with a
50µm-square pixel (detector) in response to 300K blackbody radiation[44][45]. One
of the drawbacks of VOx microbolometer is its non-crystalline structure which results
in large 1/f noises. Low NETD of 30 mK and D* of about 108 cmHz1/2/W have been
achieved for micromachined VOx microbolometers[46][47].
Metal film microbolometers, used in this study, have a positive TCR of about 0.3%/K.
Such devices normally have the detectivities in the order of 1×108 cmHz1/2/W under
room temperature operation [14]. They have comparatively low temperature
coefficient, but low noise and high lone-term stability in comparison with vanadium
oxide detectors[16]. Titanium is one of the metal materials used for microbolometers,
and reported to have the NETD of around 0.07K with f/1 optics and the TCR of about
0.25%/K to 0.35%/K [48][49]. Double sacrificial layer surface micromachining
technology used to fabricate Ti microbolomerter further improves the fill-factor of the
FPA to around 92%[50][51]. The thermal isolation effect can also be enhanced by the
technique of supporting the absorbent membrane with a lower layer, which is
anchored to the substrate. The detectivity of 2.43×109 cmHz1/2/W was reported with
Ti microbolometer [50]. In spite of its low TCR, titanium has been successfully
employed in commercial IR FPAs.
Amorphous silicon (a-Si) is another high TCR microbolometer material. With
resistivity of up to 105 Ωcm, the TCR of a-Si:N can be as high as 6%/K. However,
high noise is observed. While at low resistivity (103 Ωcm), TCR reduces to around
17
Chapter 2.______________________________________________Literature Review
2.5%/K[52]. High performance of the amorphous silicon microbolometer have been
achieved in many investigations that show a-Si is becoming an advantageous choice
for uncooled IR detector material[53~57].
YBaCuO is a well kown material for superconducting bolometers[58][59]. However,
its possible operation at room temperature has been widely studied recently[60][61].
YBaCuO bolometers also demonstrate a low 1/f noise and low power-normalized
corner frequency, fc/I 2R, where fc corresponds to the point of the frequency spectrum
at which the voltage of the current noise equals the voltage of the Johnson
noise[60][62]. The detectivity of 1×108 cmHz1/2/W was obtained with a thermal time
constant of 5 ms.
There are several other bolometer materials that have been studied, such as the
polycrystalline Si-Ge[63~65], P-N junction diode[66~68] and single crystal silicon N-
well in the standard CMOS process[69][70]. The reported detectivities for these
materials are 2.3 ×108 cmHz1/2/W[63], 1.2 ×1010 cmHz1/2/W[68] and 1.2 ×109
cmHz1/2/W[69] respectively.
2.2.2. Microbolometer focal plane array
There are two main architectures used for developing IR focal plane arrays, namely,
the hybrid and monolithic, as shown in the Fig.2.1. Hybrid microbolometer FPAs
compound the detector arrays with the silicon readout chip via some chip-to-chip
bonding techniques, such as indium bumps or loopholes, while in monolithic structure,
detector arrays are fabricated on-chip with the readout circuits.
18
Chapter 2.______________________________________________Literature Review
Fig. 2.1. Flip-chip-hybrid (a) and monolithic (b) structures of microbolometer FPA[71]
In the hybrid design, the microbolometer detector FPA and readout circuits are
fabricated on different substrates, and bonded together through indium bump or
loophole interconnection. The indium columns form both the electrical connection
between detector and signal processor, and a thermal conduction path to the silicon.
The thermal conduction must be minimized to maximize the temperature differential
across the detector and achieve high sensitivity[72][73]. Since the detector array is on
a separate chip from the one for the circuits, it can be optimized independently of the
readout circuit. A high fill factor near 100% can be achieved[17]. Hybrid structure is
used primarily when the processes of detector and ROIC are not compatible or there
are some cost issues if fabricated both together. Hybrid uncooled microbolometer
FPAs with NETD in the range of 300mK have been reported[74][75].
Modern two-dimenstional microbolometer FPAs employed in IR imaging system
usually have large number of pixels so that individual lead-outs from each pixel are
impractical. Thus, a monolithic structure with the control circuit fabricated in the
underlying silicon has been developed. In the monolithic design, the detectors are
fabricated on a thermally isolation structure on the same silicon substrate where the
19
Chapter 2.______________________________________________Literature Review
ROIC is. The monolithic thermal isolation structure typically provides a lower
thermal conduction than that of the indium columns in the hybrid FPAs. Monolithic
FPAs with the readout circuits on the same chip is the preferred choice for
microbolometer infrared FPAs as the direct on-chip electrical connection results in the
simplified package, improved performance and high reliability[76~78].
One of early studies of the fully monolithic micromachined bolometer FPA, with
simple on-chip readout circuits and preamplifier, was reported in 1993[44]. No signal
processing circuit was integrated on this chip. However, because of the improvement
of the filling factor and the reliability in the monolithic IR FPA, the monolithic FPA
technique has been widely employed and studied in the last decade [79~81]. IR
imaging FPAs with various dimensions from 128 × 128 to 640 × 480 pixels have been
reported[82~91]. The NETD for most of these FPAs have attained around 100mK,
with the best performance being less than 20 mK[84].
2.3. Readout electronics for microbolometer FPAs
The basic functions of the readout electronics for microbolometer FPAs typically
include the bias for the microbolometer, front-end amplification and signal
conditioning, pixel addressing, digitization, and error correction circuits. The block
diagram of a ROIC operated in serial pixel addressing mode is shown in Fig.2.2. The
pixels in the array are read out serially through the pixel addressing circuits that
consists of the column switches, x (column) and y (row) shifter registers. Clock
generator generates all the control and clock signal needed in multiplexer circuits and
signal processing circuits. The signal is amplified and further processed by signal
20
Chapter 2.______________________________________________Literature Review
conditioning circuits. The output from the signal conditioning circuit is digitized and
sent to the off-chip circuitry for further processing and producing the standard video
display.
Fig. 2.2. Block diagram of the ROIC for an M × N-pixel microbolometer FPA.
2.3.1. Front-end signal conditioning circuits
The microbolometer is in essence a temperature sensing resistor whose resistance
changes with its temperature. The change of the resistance can be read out as a
voltage or a current. Thus the front-end signal conditioning circuit can be classified
as voltage- or current-mode. Since the signal from the bolometer is normally very
21
Chapter 2.______________________________________________Literature Review
weak, the front-end signal conditioning circuit is critical to the performance of the
ROIC.
2.3.1.1. Voltage-mode front-end readout circuits
The simplest voltage-mode readout method is to connect the microbolometer to a bias
resistor serially, as shown in Fig. 2.3(a). The problem in this solution is that the bias
resistor damps the signal. To avoid this, an active bias scheme with high bias voltage
was proposed in [92], as given in Fig. 2.3(b), where the bolometer is connected to the
source of a PMOS transistor and the output signal is taken from its drain with the bias
resistor as a load. Large PMOS transistor should be used to avoid the significant 1/f
noise.
To overcome the problem of high 1/f noise from the active device, a modified the
passive biasing circuit with a non-overlapping chopper amplifier was reported in [93],
which is shown in the Fig.2.4. The chopping amplifier is worked as modulation
circuit, which modulate the signal to a higher frequency band so that the DC and low
frequency measurements can be avoided. Hence, the matching requirement between
bolometer resistances is eliminated and greatly reduces the 1/f noise. The offset
caused by the mismatch among the pixels (also referred to as the fixed pattern noise)
can be pre-corrected by providing a suitable chopper reference to each pixel with a
digital-to-analog converter, as indicated in the dash line.
22
Chapter 2.______________________________________________Literature Review
Fig. 2.3. Two simplest voltage-mode readout circuits
Fig. 2.4. Chopper-amplifier voltage-mode circuit
The Wheatstone bridge structure is the most commonly used voltage-mode readout
scheme for microbolometer FPAs[94][95]. The advantage of this readout circuit is
that the common-mode signal can be removed by a differential amplifier. Either a
constant or pulse bias can be employed.
23
Chapter 2.______________________________________________Literature Review
Fig.2.5 shows a Wheatstone bridge circuit with an AC bias voltage[96][97]. The AC
bias can be either a sinusoidal or a square wave voltage with zero DC level. The
frequency of the bias voltage is much higher than the thermal rolling-off frequency of
the bolometer, 1/2τ, to provide a modulated signal for amplification. A lock-in
amplifier provides the phase sensitive detection after the output of the bridge circuit is
amplified, producing a DC output voltage proportional to the resistance difference
between bolometer 1 and 2. If the two bolometers in the bridge are perfectly matched,
the 1/f noise from the amplifiers and source-followers are eliminated by the phase
sensitive detection.
Fig. 2.5. AC-bias bridge readout circuit for bolometer
A bridge readout circuit with a capacitive load shown in Fig. 2.6 was presented in
[98]. The load capacitor is polarized by a triangular wave voltage, which produces a
square wave current. This triangular voltage is built from the square voltage, which
biases the bolometer, through an active integrator using pre-amplifier. The main
advantage of this circuit is that the use of capacitive load enables the circuit free of
Johnson noise and decreases the spikes form the high frequency bias. The bias
24
Chapter 2.______________________________________________Literature Review
parameters, such as amplitude of the bias voltage, intensity of the bias current and the
slope of the bias voltage in each half period etc., are digitally controllable to obtain
the best sensitivity and bridge equilibrium. Moreover, the high frequency AC bias
separates the signal from the low frequency noise of JFET devices, such as the 1/f
noise, through modulation and demodulation.
Fig. 2.6. Bridge readout circuit with capacitive load
2.3.1.2. Current-mode front-end readout circuits
In the current mode readout, the output signal is the current from the bolometer. This
current is normally integrated and converted to a voltage at the output of the integrator.
The integration also removes out-of-band high frequency noise. It is preferred in some
applications for its high sensitivity and good noise immunity[99~101]. A simple
current-mode readout circuit for titanium bolometer FPA was employed in Ref.[48].
MOS transmission gates are used for pixel select switches to reduce the on resistance,
as shown in Fig. 2.7.
25
Chapter 2.______________________________________________Literature Review
Fig. 2.7. A simple current-mode readout for microbolometer
In another current mode readout scheme[100], a pulse current bias is employed. The
bias scheme and the pixel amplifier circuit is shown in the Fig. 2.8. With signal sel_i,
the bias current is generated from the current source that consists of M3, Mbias and
Rbias. When signal selref is pulse low, the Vref is sampled on C2 and output of pixel
amplifier is given as:
)()0(0 bolrefref VVHVV −−= (2-1)
where H(0) is the DC gain of the pixel amplifier and Vbol is the voltage drop across the
bolometer. When the signal selref is kept to high, signals phi and phi enable the
signal from one pixel be stored on C2 and act as Vref when readout the next pixel. Thus,
the signal difference between two consecutive pixels can be enhanced by the amplifier.
With this scheme, the voltage drop on the microbolometer can be continuously
measured and a high dynamic range can be achieved. Another advantage of such
circuit architecture is that it “allows accommodation of the large signal swings due to
the self-heating effect” with the high dynamic range of the circuit.
26
Chapter 2.______________________________________________Literature Review
Fig. 2.8. Pixel amplifier and pulse current bias circuits
Another commonly used current-mode readout technique is the direct injection (DI)
circuit [102][103], which is illustrated in Fig.2.9. When it is used in the
microbolomter readout circuit, the unit cell consists of a direct injection input buffer
MOS transistor, Mb, which is used to bias and sense the current from microbolometer,
and an output source follower to provides a voltage-mode output [104]. The
integration begins with resetting the voltage of the integration capacitor to a bias
voltage. Then a current flows from the integration capacitor to bolometer. The
voltage drop on capacitor is monitored and amplified. The injection efficiency of a
DI readout circuit is defined as the ratio of the current flowing into the readout circuit
to the current from microbolometer. Thus, the input resistance of the buffer MOS
transistor should be lowered in order to obtain the high injection efficiency and better
detectivity.
27
Chapter 2.______________________________________________Literature Review
Fig. 2.9. Direct injection circuit
Fig. 2.10. Direct injection circuit with CTIA integrator.
A CMOS readout circuit based on the DI input stage was reported in [54], where a
capacitive transimpedance amplifier (CTIA) was used to integrate the current signal,
as shown in Fig. 2.10. The readout is performed when switch SR is closed. Rb is a
blind bolometer which is not subject to the IR radiation and used to suppress the
background current. One Rb is required for each column. Thus, only the IR induced
28
Chapter 2.______________________________________________Literature Review
signal current is integrated by the CTIA. Similar circuit is also implemented using
switched capacitor circuit[105].
Fig. 2.11. Shared integration cell with current-mode background suppression.
The output current of DI circuits contain the large background current, which is the
bias current flowing through the microbolometer. This will significantly degrade the
circuit dynamic range and readout performance. More recently, current-mode
background suppression circuits, as shown in Fig. 2.11, have been reported[106][107].
The current from unit cell is mirrored through a cascade current mirror. The
threshold-voltage compensated current source generates a tunable background current.
Thus, current from unit cell is subtracted by a DC background current before it is
integrated on the capacitor, Cint. When using the small biasing current and large
transconductance for transistors MC1 and MC2, the drain current of MC1 can be nearly
independent of MOS threshold voltage[108]. Therefore, the shared-integration cell,
29
Chapter 2.______________________________________________Literature Review
which can be incorporated into DI circuits, provides the background current
compensation and good immunity of the threshold-voltage non-uniformity.
A constant current buffered direct injection (CCBDI) readout circuit was
demonstrated with a 64 × 64 FPA [99]. The bolometer is biased with a constant
current source, Icc. A transconductance amplifier (Gm) is used to transfer the voltage
out of the constant-current biasing readout to the current output, and an inverting
amplifier connected between the gate of the buffer transistor and the output from the
microbolometer detector. The formed negative feedback structure can decrease the
input resistance of the buffer transistor by a factor of A, the open-loop gain of the
amplifier. The unit cell circuit with the CCBDI readout is plotted in Fig.2.12. The
shared-integration background suppression circuit was incorporated to the BDI
structure and provided background suppression and the compensation of the
threshold-voltage variations. Thus, both linearity and offset of the CCBDI circuit are
improved.
Fig. 2.12. CCBDI readout circuit
30
Chapter 2.______________________________________________Literature Review
2.3.1.3. Other front-end readout circuits
A pixel-wised A/D converter concept was first proposed in [109] and the following
research is reported in [110]. The idea is to digitize the signal at early stage and
increase the noise immunity. The readout circuit is essentially a RC oscillator with R
(Rd) being the microbolometer. The IR induced resistance change of the
microbolometer is converted to frequency. Thus, the frequency of the output signal is
proportional to the incident IR radiation, and the analog signal from the
microbolometer is digitized. The subsequent oscillation frequency detection circuit
uses a observing window and counts the number of pulses in the window[111].
Fig. 2.13. Pixel-wised A/D converter circuit
2.3.2. On-chip non-uniformity correction
Non-uniformity (pixel offset) is one of the important performance limits for the IR
FPA ROICs. It may be either due to mask, lithography or other processing variables.
In a large focal plane array, the non-uniformity from pixel to pixel, which produces
the fixed pattern noise (FPN) in the output thermal images, could be very significant.
31
Chapter 2.______________________________________________Literature Review
The fixed pattern noise correction circuit is therefore an important part of the readout
electronics.
The most common FPN cancellation circuit uses the on-chip analog to digital
converter (ADC) to digitize the FPN and store the data into some off-chip memories.
When in real infrared imaging acquisition operation, the stored FPN data is read by
the on-chip digital to analog converters (DAC) and fed back into ROIC. Then ROIC
subtracts the FPN from the normal output signal pixel by pixel. Due to the large
silicon area required for high order ADC, most of the pixel-by-pixel input offset
correction circuits have only 6-bit[112] to 8-bit[113] resolutions. Therefore, the on-
chip circuit can only perform coarse correction and an off-chip cancellation is
normally required for the fine correction.
The on-chip 6-bit pixel-by-pixel FPN correction circuit with on-chip 14-bit ADC has
been reported[112]. The ROIC obtains the measured instantaneous dynamic range of
more than 76 dB at a 6.1 MHz output data rate and 60 Hz frame rate. The power
consumption is less than 500 mW. Two-point non-uniformity correction is performed
by the off-chip electronics.
As far as thermal image quality is concerned, high-accuracy on-chip ADC and DAC
are preferred for the on-chip correction of FPNs. Delta-Sigma A/D converter is seen
as a candidate to achieve high resolution without much increase in its complexity. A
second-order multiplexed oversampling Delta-Sigma A/D converter was proposed for
FPN correction in the IR FPAs [114][115]. Results from experiments showed the
accuracy better than 12 bits could be achieved in prototype ROICs.
32
Chapter 2.______________________________________________Literature Review
Since the high order Delta-Sigma A/D converter requires the very high sampling rate,
a 16-bit on-chip A/D converter with two operation modes is shown in Fig. 2.14 [116].
One 8-bit resetable Delta-Sigma converter used to obtain the most significant bits is
combined with one 8-bit conventional successive approximation A/D for residual
voltage conversion to achieve 16-bit resolution. High linearity is obtained by the first-
order Delta-Sigma converter, while speed and resolution are enhanced by the
successive approximation conversion circuit.
Fig. 2.14. Block diagram of the two-mode 16-bit A/D converter
2.2.3. Other on-chip functions
Besides on-chip offset correction, most uncooled IR FPAs integrate the gain control
and temperature stabilizer circuits on ROICs [31,84,117]. On-chip temperature-
induced offset compensation schemes using thermally shorted bolometers, as shown
in Fig. 2.15, were reported in [118] and [119]. The thermally shorted
microbolometers or the “compensation elements” have the same temperature
coefficient as the active detectors, but are not thermally isolated or thermally shorted.
Thus there is no opaque cover needed for these compensation elements. A supply
adjustment configuration to vary the microbolometer bias for different units is
33
Chapter 2.______________________________________________Literature Review
implemented by a digital-to-analog converter that adjusts the source potential of the
common gate amplifier. A transimpedance amplifier is used for integration and
amplification of the signal.
Fig. 2.15. Method for correction for temperature-induced non-uniformities with thermally shorted microbolometer elements
An on-chip bias regulator has been reported to be integrated with uncooled IR FPA,
providing the bias drift rejection[120]. It was claimed that the bias regulator can
achieve better than 100:1 drift rejection from an external 5.5 V supply.
Another bias equalization method was presented in [121]. It can provide a unique
bias to each microbolometer element, so that allows the control of the non-uniformity
over a wider range of ROIC substrate temperatures. As shown in the Fig. 2.16, an n-
bit DAC supply the bias voltage to the microbolometer through a unit cell enable
34
Chapter 2.______________________________________________Literature Review
switch. Thus, other than a conventional two-point correction, offset and gain
correction, a three-point correction process was integrated on-chip for better signal
conditioning. This approach was expected to allow the removal of the temperature
stabilizer for the microbolometer arrays.
Fig. 2.16. Block diagram of the three-point correction scheme
More recently, smart functions of the FPA readout technologies, such as neural
network concept[122], pixel averaging[123], bad pixel substitution[124] and edge
enhancement[125][126] have been presented. Neural network mimics the function of
biological sensors using silicon IC design. Pixel averaging and bad pixels substitution
has been suggested that the bad pixels in the focal plane are replaced with the average
value of the surrounded pixels. Edge enhancement and other pattern recognition
functions not only improve the IR image quality, but also can extract the user-
interested information[127]. These techniques have further enhanced the performance
of the uncooled IR focal plane array and have potential to be applied to the uncooled
FPAs.
35
Chapter 2.______________________________________________Literature Review
2.4. Self-heating effect and cancellation techniques
In order to sense the resistance change, the microbolometer must be biased by either a
voltage or a current source. The current bias current flowing through the bolometer
generates the Joule heat which raises the temperature of the bolometer. Thus, the
bolometer temperature will depend on both IR incident power and the Joule heat. The
Joule heat due to the bias current is normally much higher than the absorbed infrared
power and must be in some way compensated before the IR induced signal can be
read out. Although studies have shown that the self-heating effect should not degrade
the long-term characteristics of microbolometer[128], the readout circuit is required to
have large dynamic range to accommodate the self-heating. Several techniques for
compensating the self-heating effect have been reported.
One of the methods proposed in [129] treats the self-heating effect as a DC offset
similar to the FPN in the focal plane array and cancels it accordingly. However, this
method does not allow a subsequent integration to be performed on the readout signal,
and hence the signal-to-noise ratio cannot be further improved.
A commonly used compensation technique is to use a “blind” (dummy) bolometer to
balance the self-heating power [30][130]. The blind bolometer is a one that is
identical to the pixel bolometer, except that it is "blocked" from IR radiation. Thus, it
is only subject to the self-heating. The signal from the blind bolometer is subtracted
from the output of the normal bolometer, as shown in Fig.2.17. Although in principle,
the self-heating can be removed, in practice, it is difficult to construct the blind
bolometer that is shielded from the incident IR radiation.
36
Chapter 2.______________________________________________Literature Review
Fig. 2.17. Front-end readout circuit with the self-heating cancellation using blind bolometer
Another self-heating compensation technique based on the dummy bolometer was
reported in Ref.[131] and [132]. Being different from the previous ones, the dummy
bolometer has the same thermal mass, but higher thermal conductivity than the
normal bolometers and no IR shielding is required. The compensation circuit is
shown in Fig.2.18. The bolometer is pulse biased.
Fig.2.18. Self-heating compensation circuits in [131] [132].
37
Chapter 2.______________________________________________Literature Review
Under the condition that the bias period (tbias) is much smaller than the thermal
constant of both dummy and bolometers, the dummy device would have little
response to the IR radiation, but its temperature drift due to self-heating is almost
same as that of the normal bolometer in a pixel. Thus, the self-heating is effectively
cancelled at the output. In this scheme, the cooling rate of dummy devices is required
to be much higher than that of the normal bolometer since it needs to be cooled down
before next pixel in the same row is read. Therefore, trade-off must be made in
selecting the thermal conductance of dummy cells. The high cooling rate requires
high thermal conductance; however, if the thermal conductance of dummy devices is
too high, the condition τ<<biast may not be valid.
Two similar off-chip compensation methods, which are using the ramp voltage to
cancel the self-heating effect, have been reported for pulse bias heating of
microbolometer detectors[133]. One assumed that the detect signal sent to the pre-
amplifier was a differential voltage between the detector output and a reference. The
other directly compensated the ramp of detector voltage due to self-heating effects.
The functional architectures of these two approaches are shown in Fig.2.19.
In these methods, an additional on-chip sensor must be used to sense the resistance
value of the microbolometer. The off-chip circuits generate the ramp voltage
corresponding to the measured resistance level. The ramp voltage produced by the
circuit in Fig.2.19(a) enables the Vref to track the bolometer voltage. Similarly, the
compensation ramp in the Fig.2.19(b) is inverted so that the bias heating response of
Vd can be cancelled. The on-chip rsense was made by the resistive material with a
temperature coefficient identical to that of bolometer, but it was not thermally isolated.
Thus, it can track the ambient (substrate) temperature.
38
Chapter 2.______________________________________________Literature Review
Fig.2.19. Off-chip self-heating compensation schemes
(a)Compensation of the difference of Vref and Vd; (b) Direct compensation of Vd.
2.5. Summary
As the performance of the IR detectors continue to improve, the IR imaging
technology is advancing toward uncooled, higher sensitivity, smaller and cheaper
systems. Studies have showed that the self-heating noise of the Ti microbolometers
significantly affects the signal-to-noise ratio of the readout circuits [131][132].
Although the function filled ROIC today can provide satisfactory performance for
some applications, few self-heating techniques have been reported. The effective
suppression of the self-heating noise can further improve the performance of readout
circuit and hence the IR imaging arrays. At the present, the available on-chip self-
heating compensation methods provide only passive cancellation, which cannot be
tuned and therefore are not effective. In this project, a simple and effective self-
heating cancellation technique is proposed and related on-chip readout circuits are
developed.
39
Chapter 3._____________ ______Electro-thermal Characteristics of Microbolometer
Chapter 3. Electro-thermal Characteristics of
Microbolometer
The microbolometer is essentially a temperature sensor built on thermally isolation
structure, which absorbs the infrared radiation. Its operation is similar to that of a
simple temperature-sensitive resistor. When receiving the incident infrared radiation,
the microbolometer is heated up by the radiant power absorbed, resulting to a change
of its resistance. In order to monitor the amount of the resistance change, the
microbolometer must be electrically biased. The bias current flowing through the
sensor during the readout period also causes the Joule heating (self-heating) of the
microbolometer. Even the bias period for each microbolometer unit is very short
comparing with the frame period, the temperature increase due to the self-heating is
much higher than that due to the room temperature input infrared power. For the
purpose of obtaining a measurable change in the bolometer’s temperature, which
results from a very small input infrared power, a good thermal isolation from the
substrate is highly desired. Thus, the heat resulting from the self-heating power
cannot be quickly dissipated. As a result, the microbolometer resistance change not
only results from the information of infrared input, but also the self-heating effect.
Because of the interactions of the input infrared and the self-heating power, which are
associated with the feedback between the microbolometer and its readout circuits, the
temperature and resistance change of the microbolometer is no longer as simple as it
seems to be. An accurate simulation model based on the detailed understanding of
these behaviours is therefore required for the design of the readout electronics. Such
40
Chapter 3.___________ ________Electro-thermal Characteristics of Microbolometer
a model is also useful for the study of the self-heating effect and cancellation
techniques.
3.1. Thermal characteristics of microbolometer
The thermal behaviour of a microbolometer is mainly determined by its thermal time
constant τ, thermal conductance G and the thermal capacitance H. The relationship
among the three thermal parameters is defined by
GH
=τ (3 - 1)
Normally the thermal time constant ranges from several millisecond to less than 20
ms. Using the parameters defined above, the heat balance equation of microbolometer,
including the self-heating effect, can be written as
)( 0 ϕε ⋅+=−+ biasPTTGdtdTH (3 – 2)
where the T0 is the ambient temperature, T is the bolometer temperature; Pbias is the
self-heating power, ε is the absorptivity coefficient of the sensor element and ϕ is the
input radiation power.
The last item in the equation (3 - 2), which can be further defined by Eq.(3-3),
represents the infrared power the sensor absorbed. It causes the heating up of
bolometer elements.
ϕε ⋅=IRP (3 - 3)
Characterization of the heat propagation by an equivalent RC electrical circuit had
41
Chapter 3._____________ ______Electro-thermal Characteristics of Microbolometer
previously been reported by Auerbach et al [134]. The resistor R and capacitor C are
used to represent the thermal conductance G and the thermal capacitance H,
respectively. Thus the thermal behaviour of the sensor and the medium can both
be simulated by standard programs, such as SPICE.
Defining the thermal resistance as R = 1 / G, and with the help of current generators,
the quantities in the equation (3-2) can be well characterized by their electrical
equivalents listed in the Table 3.1, where the thermal resistance and capacitance are
lumped values, but not cover the nature of the heat distributed in the device.
Table 3.1 Electrical equivalents of the thermal behavior quantities of microbolometer
Thermal Behaviour Electrical equivalent
Sensor Temperature T (K) Voltage T (V)
Ambient Temperature T0 (K) Voltage T0 (V)
Thermal Capacitance H (J/K) Capacitance CT (F)
Thermal Resistance R (K/W) Resistance RT (Ω)
Self-heating Power Pbias (W) Current Pbias (A)
Input Infrared Power PIR (W) Current PIR (A)
Now the heat balance equation (3-2) can be rewritten using these equivalents in
electrical domain.
)( 0
IRbiasT
T PPR
TTdtdTC +=
−+ (3 – 4)
3.2. Construction of SPICE electro-thermal model of microbolometer
In order to effectively study the thermal behaviour of the microbolometer, the SPICE
42
Chapter 3.___________ ________Electro-thermal Characteristics of Microbolometer
electro-thermal model of the microbolometer is constructed and described in this
section.
3.2.1. Elector-thermal model of microbolometer
The sensor material used in this study is titanium. The temperature coefficient of
resistance (TCR) for the titanium thin film bolometer can be as high as 0.25% K-1[48].
For the metallic bolometer materials, the resistance has a linear dependence on the
temperature according to the first order approximation of TCR. The linear equation
relating resistance and temperature change is:
])([)( 000 TtTRRtR −⋅⋅+= α (3 – 5)
where α is the TCR of the sensor material, and R0 is the bolometer resistance at the
room temperature T0. The resistance, R(t), and the temperature, T(t), of the bolometer
are all time dependent.
If I(t) is defined as the bias current applied to the bolometer during readout, the bias
power (self-heating power) can be expressed as
))(()(
)()(
)()(
000
2
2
2
TtTRRtV
tRtV
tRtIPbias
−⋅⋅+=
=
=
α
(3 – 6)
where V(t) is the voltage across the resistive sensor, and
))(()(
)()()(
000 TtTRRtV
tRtVtI
−⋅⋅+=
=
α
(3 – 7)
43
Chapter 3._____________ ______Electro-thermal Characteristics of Microbolometer
When there is a bias current flowing through the microbolometer, the microbolometer
temperature increases not only due to the presence of PIR but also the effect of Pbias.
At the same time, the microbolometer resistance increases according to equation (3-5).
If the readout circuit uses a stable power supply during readout period, the current
flowing through the sensor decreases due to the increase of the resistance. Thus,
strictly speaking, I(t) should also be a time dependent variable. For the same reason
the voltage across the sensor element, V(t), also varies in the readout period. All of
these should be represented in the electro-thermal model of the microbolometer.
Combining the equation (3-4), (3-5), (3-6) and (3-7), the completed electro-thermal
model of the microbolometer can be constructed using SPICE circuit elements. The
schematic of the SPICE model is shown in the Fig. 3.1.
Fig.3.1. Electro-thermal model of microbolometer
In this model, the substrate temperature is implemented by a DC voltage source.
44
Chapter 3.___________ ________Electro-thermal Characteristics of Microbolometer
Bolom1 and Bolom2 are the two terminals of the microbolometer. The equations of
two controllable current sources, Pbias and Ibolometer(t), are defined by Eqs. (3-6) and (3-
7), respectively. The current source PIR has a constant current whose value is equal to
the input infrared power.
3.2.2. Verification of the SPICE microbolometer model
In order to verify the SPICE model of microbolometer, the experimental data reported
in Ref.[95] were used. The circuit configuration of in [95] was a Wheatstone bridge.
The experimental set-up including the pulse generator and other measurement
equipments is illustrated in the Fig.3.2.
R0 R0
PulseGenerator
Pre-Amplifier
RB R0
Vout
Digitaloscilloscope
Fig.3.2. Experimental setup in Ref.[95], where RB is the microbolometer and R0s are the fixed reference resistors.
45
Chapter 3._____________ ______Electro-thermal Characteristics of Microbolometer
In the experiments, the three constant value resistors in the Wheatstone bridge have
the very similar value as the room temperature resistance of the tested bolometer. The
magnitude of the squared pulse was 1 Volt and the duration was 60 ms. The
parameters of two single microbolometers were extracted by measuring the time
dependent output voltage of a balanced Wheatstone bridge containing a
microbolometer under pulse bias condition. Thermal capacitances and thermal
conductances of microbolometer 1 and 2 had been independently obtained using the
slope of the linear region and the saturated value of the output voltage, respectively.
To verify the SPICE model of the microbolometer, the similar Wheatstone bridge
structure was used in the simulations for the comparison with the experimental results
in [95]. The same voltage source and the resistor values reported in [95] have been
used in simulation. The equivalent parameters of our SPICE model used in the
simulations were also calculated to match those of microbolometers used in [95] at
most. The thermal resistance is calculated using the following equation.
GRT
1=
(3-8)
The thermal capacitance was calculated from
τ×= GCT (3-9)
The comparison between the simulation results and the experimental data in [95] is
done for two microbolometers. The parameters of the two microbolometers used in
both simulation and experiment are listed in Table 3.2, respectively. Figure.3.3 shows
the simulation results of the output voltage Vout together with the experimental data,
where the red curves correspond to the simulation results with the proposed SPICE
model as detectors and the black curves correspond to the experimental results
46
Chapter 3.___________ ________Electro-thermal Characteristics of Microbolometer
reported in Ref.[95]. A good agreement between the simulated and experimental
results is obtained, which indicates the correctness of the SPICE electro-thermal
model for the microbolometer.
Table 3.2. Parameters of the two bolometers used in the experiments and simulation.
Microbolometer 1 Microbolometer 2
In experiment In simulation In experiment In simulationRoom temperature
resistance 3.9 kΩ 3.9 kΩ 3.9 kΩ 3.9 kΩ
Thermal conductance 6.9×10-7 W/K 3.6×10-6 W/K
Thermal time constant 8.7ms 2.5ms
Thermal resistance 1.45 MΩ 278 kΩ Thermal
capacitance 6.0 nf 9.0 nf
TCR of the material 0.25% 0.25% 0.25% 0.25%
Fig.3.3. Comparison of simulation results using the SPICE model with the
experimental data.
47
Chapter 3._____________ ______Electro-thermal Characteristics of Microbolometer
3.3. Study of response of the microbolometer to the infrared and self-
heating power
Theoretically, the expression of the temperature variation ∆T of microbolometer can
be derived from the heat balance equation, Eq.(3-2), provided that the initial
conditions are known. When studying the microbolometer self-heating effect in the
short bias period, the input IR power can be set to zero. Then the expression of ∆T
due to the self-heating as the function of time t can be derived as:
)1(0
τt
biasP
eG
PTIR
−
=−=∆ (3 – 10)
Under normal operation, the readout period, tbias, for each microbolometer in an IR
FPA (typically several microseconds) is much shorter than the thermal time constant,
τ, of the microbolometer[135]. Thus, the equation (3-10) can be further simplified to
)( ττ
<<=⋅×
=∆ biasbiasbias ttH
Pt
GP
T (3 – 11)
The Eq.(3-11) indicates that the temperature difference due to the self-heating power
is independent of the thermal conductance of the microbolometer under the condition
of τ<<tbias .
The self-heating behaviours of microbolometer were simulated by using the electro-
thermal model in SPICE to represent the microbolometers. In an IR focal plane array
(FPA), each bolometer elements acts as one pixel. Pixels are addressed one by one so
that the signal of all pixels can be read separately during one frame period. In order to
48
Chapter 3.___________ ________Electro-thermal Characteristics of Microbolometer
understand the behaviour of the microbolometer, we first simplify the study by
considering a single microbolometer pixel, and studying its self-heating behaviours
under Wheatstone bridge readout, constant current bias readout and constant voltage
bias readout circuits.
First, the microbolometer model was connected to a Wheatstone bridge readout circuit
as shown in Fig.3.4. Rd, the detector, is replaced by its SPICE model in succeeding
simulations. R1 is a normal resistor whose resistance is equal to the room temperature
resistance of bolometer, Rd0. R0 and R2 are also normal resistors whose values are
chosen to only one tenth of Rd0 so that it is possible to fabricate them inside the pixel
near micorbolometers to reduce the fixed pattern noise. If there is no bias voltage or
current and no infrared incident applied to the bolometer detector, the bolometer
resistance would remain as Rd0. The output of the Wheatstone bridge is zero. When a
pulse voltage is applied to the bridge circuit, the Rd increases due to both the infrared
incident and self-heating power, and an output voltage Vout across the Wheatstone
bridge is thus generated.
A square pulse voltage source is applied to the Wheatstone bridge circuit to provide
the necessary bias to the microbolometer detector. The magnitude of the pulse voltage
is set to 5 V and the duration is 10 µs. Rd0 and R1 are set to 5 kΩ, and R0 and R2 are
equal to 500 Ω. The parameters of the microbolometer IR detector used in the
simulation, which provided by the Ti microbolometer development institute, Institute
of Microelectronics, Singapore, are listed in Table 3.3.
The thermal time constant of the microbolometer causes its temperature to rise
exponentially with time when the microbolometer is exposed under a constant power
49
Chapter 3._____________ ______Electro-thermal Characteristics of Microbolometer
IR radiation. This is analogous to the charging of a capacitor in an RC circuit. The
larger the thermal time constant, the slower the response of the microbolometer. The
responses of the microbolometer to the different input infrared power have been
studied. The infrared powers are set to 3.7 nW, 28.2 nW and 108 nW, respectively.
Table 3.3. Parameters of the microbolometer used in the simulation.
Microbolometer parameters Values
Detector resistance at T0, Rd0 5000 Ω
TCR, α 0.2%
Thermal Resistance, RT 2.1 × 106 Ω
Thermal Capacitance, CT 1.7 nf
Ambient temperature, T0 300°K
Initial temperature difference ∆T0 0°K
Fig.3.4. Wheatstone bridge readout circuit.
The simulation results shown in the Fig.3.5 illustrate the temperature changes of the
microbolometer at the different input powers, while the bias heating power is set to
zero. The exponential relationship between ∆T of the microbolometer and time can be
50
Chapter 3.___________ ________Electro-thermal Characteristics of Microbolometer
clearly observed. This agrees with what is predicted in Eq. (3-10). The simulation
time has been set to long enough to observe the saturation of the temperature
difference. It can be seen that ∆T rises to the saturated value after about 30
milliseconds and almost had no change after that even though the input power was
continually applied to the sensor. The saturated value reflects the input infrared
power and is sensed by the microbolometer. The temperature change of the pixel is
converted to a voltage through the microbolometer during the readout.
Fig. 3.5. Temperature changes of the microbolometer, ∆T, with different IR powers.
When a thermal image is taken by the infrared imager, the IR incident emitted by the
object will first be focused on the infrared detector array by an optical lens. Thus, the
detector array is heated up by the input infrared power for a time long enough for
sensors to reach the saturated temperature before the readout electronics begins to
work. To mimic the real operating condition, a short pulse bias voltage is applied to
51
Chapter 3._____________ ______Electro-thermal Characteristics of Microbolometer
the Wheatstone bridge circuit 50 ms after the input infrared power is applied to ensure
that the response of the microbolometer to the input power PIR is stabilized. In this
way, both self-heating and the IR response of the microbolometer can be observed.
The simulation results in the bias period of 10 µs after 50 ms are shown in Fig. 3.6.
Fig. 3.6. Output voltage (green curves) and the temperature change (pink curves) of
the microbolometer with different input IR powers.
It can be see that the in the short bias period, the bridge output and the temperature
variation of microbolometer rise almost linearly, which agree with Eq.(3-11). In the
simulation, the bolometer is first exposed to the IR radiation and lets its response
reach the saturated level before a pulse bias is applied. This is reflected by the small
voltage or temperature steps at the beginning of the bias period. The differences
between curves represent the response to the incident IR power, while the linear rise
of the microbolometer temperature or voltage is due to the self-heating effect.
52
Chapter 3.___________ ________Electro-thermal Characteristics of Microbolometer
Normally the infrared power radiated by an object around room temperature and
absorbed by the microbolometer in the pixel is below 10 nW. Therefore, from the Fig.
3.6, it can be clearly seen the output generated by the self-heating power is much
higher than that due to the input infrared power. For instance, the small step voltage
in response to the 3.7 nW infrared input power is only about 8 µV, while the output
due to the self-heating can be as high as 18 mV. Even when the infrared input power
is increased to 108 nW, the output voltage due to IR power is only about 0.6 mV.
When using Wheatstone bridge readout circuit in Fig. 3.4, assuming the bridge is
balanced, the room temperature object (PIR = 3.7 nW) can provide the output voltage
change
biasbiasdd
dout VV
RTRRRRRT
V 6
0000
00 1066.0))((
−×≈+∆++
⋅∆=∆
αα
Thus, the bias voltage required for a 3 µV output, which is minimum readable signal,
is 4.55 V. If the bias voltage is lower, objects in room temperature would be difficult
to be sensed. However, if the bias is too high, there would be large self-heating
problem. Since the system supply is 5 V, thus, the bias voltage of 5V is selected.
If the bias is applied for the duration of thermal time constant (~ ms) the temperature
of a microbolometer reaches nearly the maximum. This can be relatively large
(hundreds of degrees) in particular smaller thermal conductance used in
microbolometers. Thus, it is important to minimize the bias time which in turn can
increase the noise due to increase in bandwidth. The Johnson noise of a 5000 Ω
micbolometer can be expressed as
BkTRV bnb 42_ = (3-12)
53
Chapter 3._____________ ______Electro-thermal Characteristics of Microbolometer
where k is the Boltzmann constant that equals to1.38 × 10-23 J/K, B is the bandwidth.
According to the Eq. (3-12), to achieve a signal-to-noise (S/N) ratio higher than 1 for
a 3 µV IR signal output at 300 °K, the maximum bandwidth can be reached is:
7.10850003001038.14
)3(4
)3(23
22
≈××××
=< −
µµ
bkTRVB kHz
Thus, the minimum bias period required for 3 µV signal can be obtained as
stbias µ2.9)107.108/(1 3min_ ≈×=
In this thesis, the bias time was set to around 10 µs to avoid the over heating problem
and satisfy the S/N ratio requirement.
The self-heating effects under constant current and voltage bias readout circuits have
also been studied, respectively. Similar to using Wheatstone bridge, the bolometer is
first exposed to the IR radiation till saturated then the pulse bias is applied. Fig. 3.7
and 3.8 show the output of the readout circuits and microbolometer temperature
changes at constant current (500 µA) and constant voltage (2.5 V) bias, respectively.
It is observed that in these two biasing mode, the self-heating effects of the
microbolometer are also much higher than the output due to IR signals. In constant
current bias, the output voltage in response to the 3.7 nW infrared input power is only
about 32 µV, while the output due to the self-heating can be as high as 37.4 mV.
Even when the infrared input power is increased to 108 nW, the output voltage due to
IR power is only about 2.6 mV. Similar results can be obtained from the Fig. 3.8, the
output current due to the self-heating (7.1 µA) is more than 10 times higher than that
due to the 108 nW IR input (0.56µA).
54
Chapter 3.___________ ________Electro-thermal Characteristics of Microbolometer
Fig. 3.7. Output voltage (green curves) and the temperature change (pink curves) of
the microbolometer with different input IR powers, at constant current bias.
Fig. 3.8. Output current (green curves) and the temperature change (pink curves) of
the microbolometer with different input IR powers, at constant voltage bias.
55
Chapter 3._____________ ______Electro-thermal Characteristics of Microbolometer
3.4. Summary
In this chapter, a SPICE electro-thermal model of microbolometer is developed based
on the equivalence between the thermal and electrical domain. This model was
validated by the comparison between the simulation and the experimental results.
With the help of the SPICE model, the thermal characteristics of the microbolometer,
especially the self-heat effects, have been studied using a Wheatstone bridge readout
circuit. The study shows that the self-heating effect of microbolometer is independent
of the thermal conductance in short readout period. The simulation results of the
Wheatstone bridge, constant current bias and constant voltage bias readout circuits
have proven that the response of the microbolometer due to the infrared radiation is
much smaller than that of the self-heating. Hence the self-heating cancellation is
indispensable for the high-performance readout electronics for the IR FPA.
56
Chapter 4____Readout IC for Microbolometer with Voltage-mode Self-heating Cancellation
Chapter 4. Readout IC for Microbolometer with Voltage-
mode Self-heating Cancellation
It can be clearly seen from the description in the chapter 3 that the bolometer IR
detectors are strongly influenced by the self-heating effect. Such an effect results in a
large thermal drift in the circuits and a large portion of the dynamic range of the
subsequent pre-amplifier is therefore sacrificed. Thus, an on-chip self-heating
cancellation is crucial and necessary for the readout circuits.
Another severe problem for IR FPA application is spatial non-uniformity, which
results in a large fixed pattern noise. It will greatly degrade the dynamic range of the
readout circuit. In order to improve the performance of the FPA, the FPN must be
removed before the signal is further amplified.
In this chapter, a prototype voltage-mode readout circuit with a new on-chip self-
heating and FPN cancellation techniques is presented. The self-heating cancellation
circuit is based on the equivalence between the electrical and thermal systems and
utilizes a capacitor to mimic the thermal capacitance of the bolometer. This enables
the bias-heating to be replicated and later used to cancel itself. FPN correction is
achieved by first digitally storing the pixel-to-pixel variation under the dark condition
to an off-chip memory, and later subtracting them from the signal.
4.1. A new self-heating cancellation method
57
Chapter 4____Readout IC for Microbolometer with Voltage-mode Self-heating Cancellation
Since the thermal behaviour of the microbolometer is very similar to the charge and
discharge of a capacitor, a new self-heating cancellation circuit based on the
equivalence between the electrical and thermal capacitance is presented[136]. The
circuit uses a current source and a capacitor to mimic and cancel the self-heating
effect. The simulation result shows that the proposed circuit can effectively cancel the
self-heating and is suitable for the implementation of monolithic microbolometer-
based infrared focal plane arrays.
4.1.1. Self-heating cancellation circuit
It has been shown in the chapter 3 that the self-heating is directly dependent on the
thermal capacitance provided that τ<<biast . The smaller the thermal capacitance,
the higher the self-heating effect. Based on the equivalence between the electrical
and thermal capacitance, a replica of self-heating effect can be generated on an
electrical capacitance and used to cancel the self-heating from the microbolometer.
A self-heating cancellation circuit based on this concept is therefore
designed and schematically shown in Fig.4.1. In this modified bridge circuit,
Rd represents bolometer detector and R2 is the fixed value resistor whose
resistance is same as the room temperature resistance of Rd, that is, Rd0. The two
resistors R1 and R3 only have one tenth of the value of R2 so that R1 can be
fabricated in each pixel cell using the same material as the detector Rd. Thus, ratio
of Rd0/R1 can be made near constant within whole readout IC area. The ratio of
R2/R3 is chosen to be equal to Rd0/R1. The switches are controlled by a two-phase
non-overlapping clock whose waveform is shown in Fig. 4.2. The bridge is biased
by the constant voltage supply Vdd. However the row and column select switches
58
Chapter 4____Readout IC for Microbolometer with Voltage-mode Self-heating Cancellation
SW1 and SW2 enable each microbolometer to be pulse biased to avoid overheated.
Fig.4.1. Schematics of the self-heating cancelling scheme.
Fig.4.2. Waveforms of the clock signals.
When clock φ1 is high, switch SW3 and SW5 are closed. The capacitor is pre-charged
to a voltage equal to that at the point B, VB. The inputs of the amplifier are shorted
together by SW5, making Vd to be zero. When clock φ2 becomes high and φ1 low,
switch SW1, SW2 and SW4 are closed, and the microbolometer Rd is biased. At the
moment when SW1 turns on, VA should be equal to VB, assuming that the Wheatstone
bridge is balanced and there is no infrared incident power. As soon as SW3 turns off
and SW4 turns on, the capacitor will be discharged through I0. With correctly chosen
59
Chapter 4____Readout IC for Microbolometer with Voltage-mode Self-heating Cancellation
values for C0 and I0, a replica of the self-heating can be generated on the capacitor,
C0. This replica can be used to effectively cancel the self-heating from the
microbolometer at the input of the pre-amplifier.
4.1.2. Determination of the capacitor and current values
In order to determine values of the I-C pair that imitates the self-heating effect in the
readout circuits, the behaviour of VA in the bias period when the detector is shielded
from the IR radiation must be studied. For the purpose of simplification, only the half-
bridge structure is considered and the infrared incident power, PIR, is set to zero.
Fig. 4.3.(a) Half circuit of the Wheatstone bridge containing the microbolometer and
(b) I-C pair that generates the replica of the self-heating.
From Fig. 4.3(a), the bias heating power of the microbolometer Rd can be obtained as
201
02
21
2
))1(()1(
))(()(
TRRRTRV
tRRRtRV
Pdsw
ddd
dsw
dddbias ∆+++
∆+=
++=
αα (4-1)
60
Chapter 4____Readout IC for Microbolometer with Voltage-mode Self-heating Cancellation
where Vdd is a bias voltage across the Wheatstone bridge and RSW is the On-resistance
of the pixel-selecting switch. Since the condition α∆T << 1 is normally valid for a
metallic microbolometer, the bias power can be simplified to:
201
02
)( dsw
dddbias RRR
RVP
++≈ (4-2)
Under normal operation, the readout period, tbias, for each microbolometer in an IR
focal plane array (typically several microseconds) is much shorter than the thermal
time constant, τ, of the microbolometer. Thus, according to Eq.(3-11), the
temperature change of the microbolometer is given as the follow,
tH
PtT bias=∆ )( )( τ<<biast (4-3)
Thus, VA can be written as
tH
PRRR
RRVV
TRRRRR
tTRRVRRR
RV
RRRRRRtTRRV
RRRRV
RRRtTRRRR
RRRRV
RRRRV
tV
bias
dSW
dddA
dddSW
ddd
dSW
dd
dSWdSW
ddd
dSW
dd
dSW
ddSW
dSW
dd
dSW
ddA
⋅⋅++
−=
<<∆≈++∆
−++
≈
++++∆
−++
=
++∆−++
⋅++
=
++=
201
01
0201
01
01
1
011
01
01
1
01
01
1
1
1
1
)()0(
)1,( )()(
))(()(
)(
)(
α
αα
α
α
(4-4)
From Fig.4.3(b), Vc is equal to
tCI
VC
tQVtV ccc ⋅−=−=0
0
0
)0()()0()( (4-5)
In the self-heating cancellation circuit shown in Fig.4.1, the capacitor C0 is pre-
charged to an initial voltage Vc(0), whose value is equal to VB. Since VB is equal to
61
Chapter 4____Readout IC for Microbolometer with Voltage-mode Self-heating Cancellation
VA(0) when SW1 turns on, assuming the Wheatstone bridge is balanced. Thus,
)0()0( cA VV = (4-6)
From Eqs.(4-4), (4-5) and (4-6), it can be seen that to generate a replica of the self-
heating on the capacitor, that is to make VC(t) equal to VA(t), the following
relationship must be satisfied:
401
31
20
201
01
0
0
)(
)( dSW
dddbias
dSW
ddd
RRRHVRR
HP
RRRRRV
CI
++=⋅
++=
αα (4-7)
In order to verify the equation (4-7), the two circuits in Fig. 4.3 are simulated using
SPICE, where R1 = 500 Ω; RSW = 500 Ω; Rd0 = 5 kΩ; Vdd = 5 V. Other parameters of
the microbolometer were selected according to Table 3.3. The initial voltage at VC
was manually set equal to VA(0). Substituting the above conditions into Eq.(4-7), the
ratio 00 CI should be equal to 7.092×102 (A/F). Considering the practical realization
of the capacitor on a VLSI chip, the capacitance value was selected to be 20pf. Then
the current, I0, was calculated to be 14.2nA.
Fig. 4.4. Mimic the self-heating effect using an I-C pair.
62
Chapter 4____Readout IC for Microbolometer with Voltage-mode Self-heating Cancellation
The simulation result for VA(t) and VC(t) during the bias period are plotted in Fig.4.4.
It indicates that a replica of the self-heating effect can be generated using a current
source and capacitor pair. There is a small discrepancy between the curve VA(t) and
VC(t) when 00 CI is selected according to Eq.(4-7). This is due to the assumption,
α∆T << 1, made in the derivation of equation (4-7), while the simulation result is
exact. Such discrepancy can be eliminated by slightly tuning the ratio 00 CI by
changing the current I0. After adjusting the ratio ( 00 CI = 6.92×102 A/F), VA(t) is
perfectly resembled by VC(t). A replica of the self-heating is therefore generated.
4.1.3. Functional verification
The functionality of the self-heating cancellation circuit shown in Fig. 4.1 is verified
by SPICE simulation where the ideal switch model is used for all the switches and a
voltage control voltage source (VCVS) for the pre-amplifier. The IR power is set to
zero so that the detector output voltage is only dependent on the self-heating effect.
The voltage gain of the VCVS is selected to be unity. The values of I0 and C0 are
selected according to Eq.(4-7) with the proper adjustment. The simulation results
shown in the Fig. 4.5 indicate that the self-heating effect in the output, Vout, is
significantly reduced.
In a large microbolometer focal plane array, there are always mismatches in the
resistance and thermal parameters among different pixels. Such mismatch errors may
degrade the performance of the self-heating cancellation circuit. Further investigation
has shown that for the proposed circuit, ± 5% mismatch in microbolometer resistance
63
Chapter 4____Readout IC for Microbolometer with Voltage-mode Self-heating Cancellation
and thermal capacitance can only cause a maximum error of ±1.2 mV and ±0.8 mV in
Vout, respectively, at the end of the readout period. Thus, compared with results in
Fig. 4.5, the self-heating effect is still reduced more than 10 times. Hence, the
requirement for the dynamic range of the subsequent circuits can be relaxed.
Fig. 4.5. Simulation results for the self-heating cancellation circuit.
4.2. Self-heating Cancellation circuit design
4.2.1. Circuit design
The circuit shown in the Fig.4.1 can be realized using the standard CMOS IC
processing technology. The ideal switches can be implemented by MOS transistors.
The schematic of the self-heating cancellation circuit is given in the Fig.4.6.
64
Chapter 4____Readout IC for Microbolometer with Voltage-mode Self-heating Cancellation
Because the switch M1 is directly connected to the bias voltage source Vdd, p-MOS
transistor is used to avoid the body effect. Furthermore, in order to balance the ON
resistance of the non-ideal switch M1, switch M0 with the same type and aspect ratio
as M1 is inserted between R2 and Vdd. All other switches are implemented by n-
MOS transistors. A low-noise CMOS operational amplifier is used to implement the
pre-amplifier, which will be described in the next section.
Fig. 4.6. The schematics of the tuneable self-heating cancellation circuit.
The current sink I0 is implemented by a voltage-controlled cascode current mirror to
achieve the high output impedance and accuracy. The cascode current mirror is
tuneable through an external pin so that the ratio 00 CI can be easily adjusted to
ensure the complete bias-heating cancellation. Thus the discrepancy arising from the
approximation made in the derivation of Eq.(4-7) and the accuracy of the thermal
parameters is not important.
65
Chapter 4____Readout IC for Microbolometer with Voltage-mode Self-heating Cancellation
The output of the self-heating cancellation circuit is shown in the Fig. 4.7 with the
gain of the preamplifier is set to be unity and infrared input is set be zero. The signal
path from the fixed patter noise (FPN) correction circuit is disconnected. The
cancellation of the self-heating can be clearly seen at the output. However, there is a
small offset, about -2 mV, appearing at the output of the pre-amplifier. This offset
can also be observed at the input of the preamplifier and is found to be caused by the
non-ideal switches in the circuit. To achieve good self-heating cancellation, the sizes
of these non-ideal switches must be optimized.
Fig. 4.7. Output of the implemented self-heating cancellation circuit with non-ideal
switches.
4.2.2. Charge injection and clock feed-through
As mentioned in the last section, an offset is observed at the input of the preamplifier.
Since the gate current of the MOS transistor is zero, there should be no current flow
through the switch M2 an M3 to preamplifier. Thus, the offset cannot be due to the
unequal resistance of the transistor M2 and M3. The size of M3 does affect the pre-
66
Chapter 4____Readout IC for Microbolometer with Voltage-mode Self-heating Cancellation
charging time of C0. The large switch has a low ON-resistance, hence a small time
constant. The post-layout simulation has showed that the pre-charging time will less
than 0.1 µs when the NMOS switch M3 has the aspect ratio of as small as
2.6µm/1µm. This would have little effect to the frame rate of the self-heating
cancellation circuit.
The other possible error source is the charge injection and clock feed-through since
there are several switches connected to the input of the pre-amplifier. By observation,
there are two switches, M2 and M5, connected to the inverting input of the
preamplifier. They are controlled by the opposite clock signals. Thus, if both
transistors have the same size, the clock feed-through and charge injection should be
cancelled[137]. At the non-inverting input, there are three switches. M3 and M5 are
controlled by a signal opposite to that on the gate of M4. Thus, the charge injection
caused by two switches, M3 and M5, to this node is opposite to that of M4. When the
size of M4 is approximately equal to the sum of that of M3 and M5, the charge
injection should be cancelled.
The above analyses have been proved by the simulation as shown in the Fig. 4.9,
where M3 and M5 have same size, while M4 is about twice of the M3 or M5. That is
5,34 MM LWK
LW
⎟⎠⎞
⎜⎝⎛=⎟
⎠⎞
⎜⎝⎛ (4-8)
where . The charge injection is effectively compensated. In simulation, the
infrared input power is set to zero, the aspect ratios of two PMOS switches were set to
20 µm/1µm, and the NMOS switch M2 is chosen to be the same size as M3. The bias
period of the microbolometer is set to 10 µs.
3.2≈K
67
Chapter 4____Readout IC for Microbolometer with Voltage-mode Self-heating Cancellation
Fig. 4.8. Charge injection vs. different aspect ratio of M4. The upper plot shows non-inverting inputs of the preamplifier, while the lower plot shows differential inputs of
the preamplifier, at the different aspect ratio of M4.
Fig. 4.9. Charge injection after the size of M4 is optimized. The upper plot shows
non-inverting inputs of the preamplifier, while the lower plot shows differential inputs of the preamplifier, when size of M4 over M3, M5 are around 2.3.
68
Chapter 4____Readout IC for Microbolometer with Voltage-mode Self-heating Cancellation
4.2.3. Final simulation results
The overall self-heating cancellation simulation result is shown in Fig. 4.10. The
circuit parameters used in the simulation are given in the Table 4.1. The room
temperature resistances of microbolometer are selected according to the available two
types of the single bolometers fabricated by Institute of Mircoelectronics, Singapore.
The thermal parameters of the microbolometers are those from Table 3.3. The gain of
the pre-amplifier is 5. The value of I0/C0 is calculated using Eq.(4-7), and slightly
adjusted to achieve better self-heating cancellation performance. It can be seen from
Fig. 4.10 the self-heating effect can be effectively cancelled and there is no offset.
Table 4.1. The circuit parameters after optimization
Circuit parameters Values
Room temperature resistance of microbolometer 1 5 kΩ
Room temperature resistance of microbolometer 2 2.6 kΩ
Aspect ratio of M0 and M1 20 µm/1µm
Aspect ratio of M2, M3 and M5 2.6 µm/1µm
Aspect ratio of M4 6 µm/1µm
I0, for microbolometer1 13.8 nA
I0, for microbolometer2 29.6 nA
Capacitance C0 20 pf
69
Chapter 4____Readout IC for Microbolometer with Voltage-mode Self-heating Cancellation
Fig. 4.10. Simulation results of optimized self-heating cancellation circuit.
4.3. Incorporation the voltage-mode self-heating cancellation into ROIC
In order to demonstrate the suitability of the self-heating cancellation scheme for the
focal plane array, a readout IC for a 128 × 128 microbolometer FPA is designed. The
architecture of the ROIC is shown in Fig. 4.11. Each pixel in the FPA comprises all
the elements enclosed by the dash line in Fig. 4.6. Partially parallel readout structure
is adopted, in which 16 columns are read out simultaneously to achieve longer signal
integration time. A long integration time is essential for the low-noise design. To
satisfy the requirement of most commercial applications, the frame rate of the
imaging system should be higher than 30 Hz. For the 128 × 128 pixel FPA, the
proposed partially parallel readout scheme has a readout duration for each pixel as
70
Chapter 4____Readout IC for Microbolometer with Voltage-mode Self-heating Cancellation
long as 16 µs for a 30 Hz frame rate, and 8 µs for 60 Hz frame rate. In the following
simulation, the pixel readout duration is set to be 10.8 µs, which gives a frame rate of
45 Hz. All the column switches are implemented by a single-NMOS transistor. The
on-chip clock generator provides all the necessary clocks for the ROIC. 128-row
digital shift register and partial 16-column shifters are used to generate the pixel
addressing signals, which drive the row and column switches, respectively. The
frequency of the row select signals is 128 times higher than that of the column select
signals so that the pixels in one column can be addressed row by row.
Fig. 4.11. Partially parallel readout architecture of the voltage-mode ROIC.
71
Chapter 4____Readout IC for Microbolometer with Voltage-mode Self-heating Cancellation
When a pixel is addressed, its output, generated by both infrared and bias heating, is
fed to the inverting input of the differential pre-amplifier, as shown in Fig. 4.12. At
the same time, the self-heating cancellation circuit generates a replica of the self-
heating voltage which is applied to the non-inverting input under the control of the
clock φ1 and φ2. Thus, the self-heating can be cancelled during the readout. In
addition, the fixed pattern noise of FPA obtained during the initial calibration period
under dark condition is also fed to the non-inverting input of pre-amplifier during
readout for FPN cancellation. The output from the pre-amplifier should therefore
contain neither self-heating effect nor FPN. The amplified infrared signal is further
integrated and digitized by an 8-bit on-chip A/D converter. The on-chip 8-bit A/D
and 8-bit D/A converters are selected from AMS (AustriaMicroSystems) analogLib
library.
Fig. 4.12. Block diagram of the fixed pattern noise correction circuit.
4.3.1. Clock and control signals generation
All the clock and the control signals are generated on-chip with an external master
clock signal. The clock generation circuit provides all the necessary clocks and
72
Chapter 4____Readout IC for Microbolometer with Voltage-mode Self-heating Cancellation
control signals for the self-heating cancellation circuit, amplifier, integrator, the on-
chip A/D and D/A converters. The block diagram of the clock generation circuit is
shown in Fig. 4.13. The main_clk generator block generates ф1, ф2, main_clk and
int_ctrl from an external input clock. The main_clk is the pixel clock for the FPA.
The row and column select clocks are generated by the addressing clock generator,
using main_clk as the input clock. The addressing clock generator also produces a
frame synchronization signal when the last pixel in the array is addressed. This
synchronization signal is send to the off-chip circuitry and the imaging acquisition
card to synchronize the operation of the entire system. A/D clock generator block
generates the ad_clk and ad_start signals for the on-chip A/D converter.
Fig. 4.13. Block diagram of clock and control signal generator.
4.3.1.1. Clock generator for main_clk and φ1, φ2
The clock generation circuit is shown in Fig. 4.14, where an external master clock,
73
Chapter 4____Readout IC for Microbolometer with Voltage-mode Self-heating Cancellation
clk, is used to derive all the internal clocks. Clock φ1 and φ2 are designed to define the
bias and amplification period. This period is one master clock cycle shorter than the
main_clk cycle so that before the microbolometer detector is biased there is enough
time to pre-charge the capacitor C0 in the self-heating cancellation circuit. The
int_ctrl clock, which is generated for the on-chip integrator, is designed in such a way
that it allows the signal integration to start at half clk cycle after the bias period
and end at half clk cycle before. Hence, the transient noise from φ1 and φ2 can be
avoided. The high period of the int_clk resets the integrator. An external reset signal
is used to reset all the counters, shift registers and flip-flops to ‘0’. The simulation
results are plotted in Fig. 4.15.
Fig. 4.14. Schematic of the pixel clock and control signal generator.
74
Chapter 4____Readout IC for Microbolometer with Voltage-mode Self-heating Cancellation
Fig.4.15. Waveforms of int_ctrl, φ2, φ1 main_clk and clk.
4.3.1.2. Clock generator for A/D converter
The A/D converter is controlled by two clock signals, ad_clk and ad_start. It requires
9 clock cycles to perform one data conversion. Among them, one cycle is for sample
and hold, and the other 8 cycles are for converting the signal to 8-bit data. When
ad_start is low, no conversion occurs. First rising edge of the ad_clk after the
ad_start goes high will reset the ADC. At the second ad_clk rising edge, input signal
is sampled and held. After the ninth clock cycle, the first data is valid at the output of
ADC. Thus, conversion will repeat every 9 cycles while the ad_start is high.
The design of the ad_clk and ad_start signal generators should ensure that the
converter is started while the first pixel is read out and the signal should be sampled
and held after a considerable period of integration. In the circuit illustrated in Fig.
75
Chapter 4____Readout IC for Microbolometer with Voltage-mode Self-heating Cancellation
4.16, the frame signal, which is logic high when the last pixel in FPA is reading, is
used to make the ad_start active at the start of the seventh cycle after the first frame.
The ad_clk is obtained by inversing the external input clock, clk, so that its second
rising edge after ad_start will be half cycle before the end of the integration period.
Fig. 4.17 shows the timing waveform of the related signals.
Fig. 4.16. Schematic of the ad_clk and ad_start signal generator.
Fig. 4.17. Timing waveforms of ad_start, ad_clk, frame, int_ctrl and main_clk.
76
Chapter 4____Readout IC for Microbolometer with Voltage-mode Self-heating Cancellation
4.3.1.3. Pixel addressing signal generation
The pixels in the FPA are addressed by the row and column select signals. The circuit
used to generate the addressing clocks is shown in Fig. 4.18. Row addressing is
accomplished by a 7-bit cyclic synchronous counter, a 7-input NOR gate and a 128-
bit shift register (y-shifter). The output of the NOR7 gate is set to high when all the
outputs of the 7-bit counter are low. Thus, for every 128 main_clk cycles will output
a pulse. The pulse propagates through the y-shifter at every rising edge of the
main_clk and generates the row select signals.
Column select signals are generated in a similar way as the row select signals. The
last bit output of the 7-bit counter, Q6, is inversed and used as the clock of the 4-
bit cyclic counter so that the output of NOR4 gate becomes high for 128 main_clk
cycles and repeats every 2048 main_clk cycles ( 16 cycles for col_clk). The 16-bit
partial x-shifter shifts the output of the NOR4 gate and generates the column select
pulses. Since there are 128 columns in the microbolometer focal plane array, eight
same partial x-shifters operate at the same time under the control of the col_clk and
the output from the NOR4 gate.
The frame pulse is obtained at the end of each frame by ANDing the last row and last
column select signals. This signal is used to control the start of the A/D conversion
and to synchronize the off-chip circuitry of the overall IR imaging acquisition system.
Fig. 4.19 and 4.20 show the timing waveforms of the first 6 row select signals and 6
column-select signals, respectively.
77
Chapter 4____Readout IC for Microbolometer with Voltage-mode Self-heating Cancellation
Fig. 4.18. Schematic of the X-Y scanning clock generator.
Fig. 4.19. Waveforms of the scanning clocks of the first 6 rows and main_clk.
78
Chapter 4____Readout IC for Microbolometer with Voltage-mode Self-heating Cancellation
Fig. 4.20. Waveforms of the scanning clocks of the first 6 columns and col_clk.
4.3.2. Pre-amplifier design
A typical pre-amplifier for microbolometer readout circuits must amplify the input
signal to a required level with very low noise, since the signal from the bolometer can
be as low as 3 µV. Furthermore, to satisfy the requirements of the 30 Hz frame rate
for a 128 × 128 microbolometer array, the gain bandwidth (GBW) of the preamplifier
should be larger than 500 kHz. For the partial parallel readout, the GBW of 100 kHz
is accepted. The differential input range of the preamplifier should cover the FPN of
the FPA. The FPN, caused by the variation of bolometer resistance, is typically
within ± 10% of the nominal value. In our design, the nominal value of
microbolometer resistance at the room temperature is 5 kΩ, which leads to the FPN of
± 100mV. The common mode range of the preamplifier can be from 0.3 V to 3.0 V.
79
Chapter 4____Readout IC for Microbolometer with Voltage-mode Self-heating Cancellation
The lower limit of the common mode range is critical because the resistors in the
lower arms of the Wheatstone bridge could be only one tenth of the microbolometer
resistance. The offset is another important parameter for preamplifier. It should be
limited to several microvolts level. Since the settling time of switching from one
pixel to another must be shorter than the φ1 period, the slew rate of the preamplifier
should also be higher than 1 V/µs if the amplifier gain is higher than 2. In summary,
the requirements of the preamplifier lead to the following specifications:
1) GBW > 0.8 MHz
2) Differential input range > ± 150 mV
3) Common mode input range from 0.2 to 3.5 V
4) Input referred offset < 10 nV (adjustable)
5) Integrated noise level from DC to 500 kHz < 3 µV;
6) Slew rate > 2 V/µs
The pre-amplifier is a two-stage operational amplifier shown in the Fig. 4.21. In
order to satisfy the design specifications, several design considerations have to be
taken into account. For the implementation of low-noise preamplifier, PMOS
differential input stage is used. Since the PMOS input transistors must be kept in
saturation region for a common-mode input level near the ground rail for single power
supply operation, the common-mode input range requirement can be easily achieved
with a PMOS input stage. The output stage is a common-source high gain stage with
low output impedance and high drive current.
To reduce the flicker noise, the gate length of 8 µm is used for the input transistor M0
and M1, and 10 µm for active load M2 and M3. Long channel length also provides
80
Chapter 4____Readout IC for Microbolometer with Voltage-mode Self-heating Cancellation
good matching characteristics and hence reduces the input offset. The differential
input stage is biased by a current source which consists of M4, M5, M6, M7, and
generates around 100 µA current to the tail of the input transistor pair. Two small
resistors R1, R2 are added between the sources of M2, M3 and ground, respectively,
for external offset adjustment. The values of R1 and R2 are designed to make the
systematic DC offset of the amplifier as low as possible. Transistor M8 ~ M12 are
added to form the push-pull output stage. The frequency response and the stability of
the amplifier are ensured by the Miller capacitor, Cc, and the nulling resistor, Rc.
Since the offset in operational amplifiers not only comes from the systematic design,
but also from the imperfect fabrication of matched devices, careful layout of the
matched devices is planed in the design to reduce process-related offset. The
common-centroid cross-coupling layout strategy together with double guard rings is
adopted for critical devices. All the devices that need to be matched were arranged in
close proximity to minimize the effect of spacing-dependent parameter mismatch.
These include the input pair, active loads and current mirror sources. The layout of
the preamplifier is shown in Fig. 4.22.
The overall post-layout performance of the preamplifier, simulated with 5 V power
supply and a 5 pf capacitor load, is summarized in the Table. 4.2. The input referred
noise is integrated over the bandwidth of the preamplifier since there is no filter in the
readout circuit.
81
Chapter 4____Readout IC for Microbolometer with Voltage-mode Self-heating Cancellation
Fig. 4.21. Schematic of the pre-amplifier.
Fig. 4.22. Layout of the preamplifier.
82
Chapter 4____Readout IC for Microbolometer with Voltage-mode Self-heating Cancellation
Table 4.2. Post-layout Simulated pre-amplifier parameters
Parameters Results
DC gain 107.3 dB
GBW 4.6 MHz
Phase margin 82 degree
Input referred offset 5.98 nV
Positive slew rate 6.4 V/µs
Negative slew rate -6.2 V/µs
Noise @ 1 kHz 32 HznV
Input referred noise integrated @ 1 Hz ~ 4.6 MHz 1.2 µV
Core preamplifier area 112 × 137 µm2
Power consumption 8 mW
4.3.3. Fixed pattern noise correction
The fixed pattern noise is caused by the spatial non-uniformity in the infrared focal
plane array. The existence of the spatial non-uniformity is due to the variation in the
fabrication process. In a large focal plane array, the non-uniformity from pixel to
pixel can be significantly larger than the signal due to incident infrared. Therefore, it
must be corrected in the readout circuit. The schematic of the FPN correction circuit
is shown in Fig. 4.23.
83
Chapter 4____Readout IC for Microbolometer with Voltage-mode Self-heating Cancellation
Fig.
4.2
3. S
chem
atic
of t
he F
PN c
orre
ctio
n ci
rcui
t.
84
Chapter 4____Readout IC for Microbolometer with Voltage-mode Self-heating Cancellation
There are two operation modes in this circuit, namely, the calibration and imaging
mode. When in calibration mode, the detector array is shielded from IR incident. The
FPN feedback path from the output of D/A converter is disconnected by the
transmission gates that is controlled by signal Calib/Imaging. The fixed pattern noise
of the focal plane array is then read and sent to the inverting input of the preamplifier.
At the same time, the replicated the self-heating signal is fed to the non-inverting
input of the pre-amplifier. Thus, the self-heating effect is removed and leaves only the
FPN at output of the pre-amplifier. The FPN is then integrated, digitized by an on-
chip 8-bit A/D converter, and stored in an off-chip memory. During the imaging
mode, the transmission gates are open. The 8-bit on-chip D/A converter reads the
FPN from the off-chip memory and sends them back to the input of the pre-amplifier.
The preamplifier operates similar to a subtracter, subtracting the self-heating noise
and fixed pattern noise from the readout signals from the FPA. Thus, both self-heating
and FPN can be cancelled at the input of the pre-amplifier. The output signal of the
pre-amplifier with a small residual FPN is then digitized for further fine off-chip
correction.
To achieve a Noise Equivalent Temperature Difference (NETD) better than 100 mK,
the FPN correction with at least 16-bit resolution is needed. In our design, only 8-bit
coarse FPN correction circuit is built on chip. The fine correction will be performed
off-chip.
The pre-amplifier described in the section 4.3.2 is adopted to implement all amplifiers
in the FPN correction circuit. The values of R1 ~ R6 are selected according to the
following equations.
85
Chapter 4____Readout IC for Microbolometer with Voltage-mode Self-heating Cancellation
DACADC GainGainGainRR
RRRRRR
××=
===
int3
5
65
43
21
1 (4-9)
One of the important factors, which must be considered in selection of the resistor
values, is that the integrator stage cannot be saturated when circuit is working in the
calibration mode. Therefore the ratio 1R5R
is chosen to around unity.
The external gain_ctrl pin controls the time constant of the integrator so that different
time constants can be set for the calibration and imaging operation, respectively, and
the full dynamic range of the integrator can be utilized for the small signal after FPN
correction.
Post-layout simulation of the circuit in Fig. 4.23 had been run with a FPN input
between pin in1 and in2 fed to the amplifier. Considering the ±100 mV variation,
which equivalent to a ± 10% non-uniformity in the microbolometer FPA, and other
variation from the circuit fabrication process, the maximum input FPN range is set to
±130 mV. A voltage source is used in simulation to generate the FPN input. The
digitized FPN is saved when circuit operates in calibration mode and the FPN
feedback from DAC is disconnect. In the imaging mode, the saved FPN data is the
input to Din<7:0> so that the residual FPN can be observed at the output of the pre-
amplifier. The simulation is repeated with the simulated FPN stepped from -130 mV
to 130 mV and results are shown in the Fig. 4.24. After the correction, the FPN can
be reduced to less than ± 400 µV for the maximum input FPN range of ±130 mV. In
the simulation, the self-heating block is disabled.
86
Chapter 4____Readout IC for Microbolometer with Voltage-mode Self-heating Cancellation
Fig. 4.24. Simulation results of fixed pattern noise correction.
4.4. Summary
A new self-heating cancellation technique for microbolometer readout circuits based
on equivalence of thermal capacitance and electrical capacitor is described. With the
proper chosen values, a current and capacitor pair can be used to generate a replica of
the self-heating. The design equations for determining the value of the current source
and capacitor have been derived. The self-heating cancellation circuit is optimized
and simulation results show that an effective self-heating cancellation can be
achieved. The tuneable feature of the cancellation has also been demonstrated.
A partial parallel voltage readout integrated circuit for microbolometer focal plane
array is proposed based on the new self-heating cancellation technique. 8-bit on-chip
A/D and D/A converters form the coarse FPN correction circuit. An 8-bit correction
has been achieved using the fixed pattern noise correction circuit. Although the ROIC
is demonstrated for a 128 × 128 microbolometer FPA, it can be easily extended to a
focal plane array with a larger dimension.
87
Chapter5. Readout IC for Microbolometer with Current-mode Self-heating Cancellation
Chapter 5. Readout IC for Microbolometer with Current-
mode Self-heating Cancellation
In currently available state-of-the-art IR FPA readout circuits, noise requirement is
more and more strict so that the system sensitivity can be further enhanced.
Comparing with the voltage-mode readout circuit, current-mode circuit shows its
good noise immunity. However, there is no self-heating cancellation circuit reported
for current-mode microbolometer readout circuits. In this chapter, a current-mode
readout circuit with both of self-heating noise cancellation and FPN correction is
presented. The concept of using the electrical equivalence of the thermal system to
generate the replica of self-heating effect in voltage-mode readout circuits is extended
to the current-mode readout circuits. The cancellation of the self-heating is
demonstrated by both simulation and experiment. An on-chip 8-bit current-steering
D/A converter is designed to perform the coarse correction of the FPN.
5.1. Current-mode self-heating cancellation circuit
The proposed current-mode self-heating cancellation circuit is based on the same
concept as that described in the voltage-mode cancellation circuit, that is, the
equivalence between electrical and thermal capacitance. The details are described in
the following sub-sections.
5.1.1. Current-mode self-heating cancellation scheme
88
Chapter5. Readout IC for Microbolometer with Current-mode Self-heating Cancellation
The schematic of the current-mode microbolometer self-heating cancellation circuit
for single microbolometer is shown in Fig. 5.1 [138]. Rd represents the
microbolometer. R1 is a fixed-value resistor whose resistance is equal to the room
temperature resistance of Rd, Rd0. M1 is used to balance the ON-resistance of M0
and is set to always on. Three unity gain cascade current mirrors, (M2 ~ M5), (M6 ~
M9) and (M10 ~ M13) form the current subtraction circuit. The output current of the
current mirror (M2 ~ M5) is set by the microbolometer Rd, while the output current of
another current mirror (M6 ~ M9), which is completely identical with the current
mirror (M2 ~ M5) to obtain the well matched structure, is set by the resistor R1. The
current output from (M10 ~ M13) is equal to current Id. Assuming the ramp current
generator generates the replica of the self-heating effect of Id, the differential output
current should be equal to the difference between Id and the sum of Iref, the DC current
of Id, and Iramp, the replica of self-heating noise, and contain only infrared signals. The
differential current is fed to a subsequent current integrator.
The circuit of the ramp current generator is depicted in Fig. 5.2. The switches are
controlled by the two-phase non-overlapping clock φ1 and φ2. When clock φ1 is high
and φ2 low, the microbolometer is not biased. At same time, M14 is on and the
capacitor C0 is pre-charged to a voltage, which is equal to the drain voltage of M6,
VD6. Controlled by the complementary signal of φ2, M15 is off. Therefore, VDS of
M17 is approximately equal to zero. While clock φ2 becomes high and φ1 low, switch
M0, M15 and M16 are on, and the bolometer Rd is biased. Under the assumption of a
metal-film bolometer, Rd increases due to the self-heating effect, resulting in the
decrease of Id. The operation amplifier in Fig. 5.2 is used to track the voltage on the
capacitor C0 and feed it to the drain of M17. Since M14 is off but M16 on, I0 is
discharging C0 so that the drain voltage of M17 is linearly decreasing while its source
89
Chapter5. Readout IC for Microbolometer with Current-mode Self-heating Cancellation
voltage remains unchanged and is equal to VD6. Note that VD6 is mainly determined
by Iref since Iramp is very small. The drain source voltage of M17, - I0/C0 t, is designed
to much lower than the threshold voltage of the PMOS transistor so that M17 is
operating in the triode region. With correctly chosen values for C0 and I0, the ramp
current can have the same gradient as the self-heating current of the microbolometer,
Id. Thus, a current-mode replica of self-heating is obtained.
Fig. 5.1. Schematic of the current-mode self-heating cancellation circuit.
Fig. 5.2. Schematic of the ramp current generator.
90
Chapter5. Readout IC for Microbolometer with Current-mode Self-heating Cancellation
5.1.2. Determination of the capacitor and current values
During readout, the resistance change of a metal-film microbolometer is equal to
Rd0α∆T. Assuming there is no IR incident input and the ∆T is only caused by the
microbolometer’s self-heating effect, ∆T can be derived when the bias period is much
shorter than the thermal time constant of the microbolometer.
tH
PtT bias=∆ )( (5-1)
For the circuit shown in Fig. 5.1, the Pbias can approximately be written as
)1()(
)(
0
20
20
TRVV
RVV
Pd
Ddd
d
Dddbias ∆+
−=
−=
α (5-2)
where Vdd is bias voltage and VD0 is the drain voltage of the M0.
The drain source current of the transistor M2 and M4, Id can be written as
1)T ( )(
)(
)(
)(
)(
since020
30
02
03
0
0
0
0
00
0
0
0
00
0
<<∆≈⋅⋅
−−≈
⋅⋅
−−
−=
⋅⋅−
−−
=
∆−⋅
−=
−=
αα
α
α
α
ddd
Dddref
dd
dDdd
d
Ddd
bias
dd
dDdd
d
Ddd
d
dd
d
Ddd
d
Dddd
RRtRHVV
I
tRRH
RVVR
VV
tH
PRR
RVVR
VV
RtTRR
RVV
RVV
tI
(5-3)
The drain source current of the transistor M6 and M8 is equal to Iref – Iramp. Assuming
current mirrors are matched, the output current of the current subtracter can be
expressed as
)( ramprefddiff IIII −−= (5-4)
91
Chapter5. Readout IC for Microbolometer with Current-mode Self-heating Cancellation
Iramp is the drain current of M17, which linearly depends on its VDS when operating in
the triode region and can be written as
tCI
VVL
WC
VtCI
VVVL
WC
VVVL
WCI
THGSoxp
DDTHGSoxp
DSTHGSoxpramp
0
017
60
0617
17
)(
)()(
)(
−=
−−−−=
⋅−−=
µ
µ
µ
(5-5)
where pµ is the hole mobility, is the gate oxide capacitance per unit area,
(
oxC
LW )17 is the aspect ratio of M17 and VTH is the threshold voltage. When there is no
infrared incident input and if the self-heating is completely cancelled, Idiff should be
equal to zero. Substituting Eqs. (5-3) and (5-5) into (5-4), and let Idiff = 0, it yields
THGSoxpd
Ddd
VVL
WCRHVV
CI
−⋅
⋅
−=
17
20
30
0
0
)(
1 )(
µ
α (5-6)
In other word, if Eq.(5-6) is satisfied, the ramp current can successfully
compensate the self-heating effect.
The current source I0 can be implemented by an adjustable current-sink so that the
gradient of the ramp current can be adjusted through an externally voltage-
controlled pin. M17 can also be replaced by a resistor, but with an area penalty.
The use of PMOS transistor M17 can provide another mean to tune the ramp
current if its gate voltage is made adjustable.
5.1.3. Functional verification
The functionality of the current-mode self-heating cancellation circuit has been
verified by SPICE programs. In simulation, Rd is replaced with a microbolometer
model. The period of the pixel clock is set to 10 µs. The bias period (ф2 period) of
92
Chapter5. Readout IC for Microbolometer with Current-mode Self-heating Cancellation
the microboloemeter detector is 9.5 µs in simulations. The results of the simulation
show that 0.5 µs is long enough to pre-charge the capacitor C0. When C0 is designed
to be 10 pf, the ф1 period can be further decreased to around 0.2 µs to increase the
bias period of detectors. The IR incident power per pixel is set to zero so that self-
heating effect and its cancellation can be clearly observed. The microbolometer
related parameters used in simulation are set as in Table 3.3. Process related
parameters in Eq.(5-6) are based on the device models from the AMS 0.6-µm CMOS
double-poly triple-metal process. Other circuit parameters are listed in Table 5.1.
Care should be taken when choosing the sizes of the switch transistors, M14 ~ M16.
The charge injection effects similar to that occurred in the voltage-mode self-heating
cancellation circuit should be avoided.
Table 5.1. Current-mode self-heating cancellation circuit parameters in simulation
Circuit parameters Values
Rd0, R1 5000 Ω
Aspect ratios of M2, M3, M6, M7 400.5µ/3µ
Aspect ratios of M4, M5, M8, M9 400µ/3µ
Aspect ratios of M10, M11 800µ/3µ
Aspect ratios of M12, M13 618µ/3µ
Aspect ratio of M14, M16 11.2µ/0.6µ
Aspect ratio of M15 12µ/0.6µ
Aspect ratio of M17 29.6µ/5µ
C0 10 pf
I0 (after tunning) 18 nA
Vdd 5 V
The simulation results in Fig. 5.3 shows that after proper cancellation, the current Idiff
is almost zero within whole bias period. The self-heating is effectively cancelled
using the proposed current-mode cancellation circuit. The ratio I0/C0 can be tuned by
93
Chapter5. Readout IC for Microbolometer with Current-mode Self-heating Cancellation
I0 through the Ramp_ctrl pin to achieve complete self-heating cancellation. Fig. 5.4
shows the tuneable feature of the ramp current generator, where I0 is tuned to be 25.2
nA, 18 nA and 10.8 nA. The slope of Iramp is therefore changed according to I0. The
tuneable feature ensures a good self-heating cancellation since errors from
assumptions made in deriving Eq.(5-6) and the variation of the process parameters are
tolerable.
Fig. 5.3. Simulation results for current-mode self-heating cancellation.
Fig. 5.4. Tuneable feature of the ramp current generator.
94
Chapter5. Readout IC for Microbolometer with Current-mode Self-heating Cancellation
5.2. Current-mode ROIC with self-heating cancellation
The readout circuit with the proposed current-mode self-heating cancellation has a
simpler structure than the one in voltage-mode when employed in an IR FPA. In
current-mode, each pixel contains a microbolometer and a MOS transistor as the
ROW selective switch. Thus, only three connections (Row, Vbias and Ipout) are needed,
compared with that in the voltage-mode where there are four global connections
(Row, Vdd, GND and Vpout). This is shown in Fig. 5.5.
Fig. 5.5. Pixel structure of the voltage-mode(a) and current-mode(b) readout circuit
A current-mode readout circuits for the M × N pixels FPA is shown in Fig. 5.6. The
digital circuitry is similar to that in the voltage-mode readout systems. The X-shifter
generates the scan control signals for N column-switches, while Y-shifter generates
the control signals for M row-switches. These switches are all implemented by the
NMOS transistors. Since it is a column-major scanning system, frequency of the
clock for Y-shifter should be M times higher than that of the clock for X-shifter. The
current output from each pixel is then multiplexed and sent to the current-readout and
self-heating cancellation circuits (circuits enclosed by the dash-line in the Fig. 5.1).
After the current-subtraction circuits, the output current Idiff is integrated and
95
Chapter5. Readout IC for Microbolometer with Current-mode Self-heating Cancellation
converted to a voltage signal, which is then, digitized by the on-chip voltage A/D
converter and saved in an off-chip memory. The Calib/Image switches the ROIC
between calibration and imaging mode. When in the calibration mode, the FPA
should be shielded from IR radiation. Assuming a perfect self-heating cancellation,
the saved digital output, should contain only the fixed pattern noise of the FPA. In the
imaging mode, this digitized FPN is converted back in the analog domain, fed to input
of the integrator and cancels the FPN. Thus, the final output should only contain the
IR signal and small residual FPN.
In addition to the clocks needed by two digital shifters, the on-chip clock generator
also provides ф1 and ф2 for readout and self-heating cancellation circuit and all other
control signals required by the integrator and A/D converter. Partial parallel readout
structure can also be adopted for the current-mode readout scheme in Fig. 5.6 if the
microbolometer FPA is a large array, so that the signal integration period longer than
5 µs for each pixel can be guaranteed.
Fig. 5.6. Readout circuit architecture of the current-mode ROIC for an M×N FPA.
96
Chapter5. Readout IC for Microbolometer with Current-mode Self-heating Cancellation
5.2.1. Digital circuitry of the current-mode ROIC
The digital blocks are similar to those in the voltage-mode ROIC, which are described
in the Chapter 4. The main_clk is the pixel clock used as the basic clock to generate
the X-Y scanning pulses. The int_ctrl signal is the pulse to control the current
integrator. When the int_ctrl is high, the integrator output is reset to the reference
voltage (2.5 V in the designed circuits) and the capacitor is quickly discharged. When
int_ctrl is low, the current Idiff charges the capacitor and generates the voltage signal.
The clock generator makes the low period of the int_crtl shorter than the bias period
of the microbolometer detector by one clk cycle so that the transient noise of the φ1, φ2
can be avoided. The X-Y address clock generation circuit is given in Fig. 5.7 where
Y-shifter has M bits and X-shifter N bits. The frame signal is the synchronization
signal for A/D clock generator and off-chip circuitry, whose falling edge indicates the
start of the readout period for the first pixel.
Fig.5.7. X-Y address clock generator for M × N FPA.
97
Chapter5. Readout IC for Microbolometer with Current-mode Self-heating Cancellation
5.2.2. Current subtraction circuit
The current subtraction circuit formed by three current mirrors is the most critical part
in the current-mode ROIC. In order to reduce the effect of channel length modulation,
cascade structure mirror, as shown in Fig. 5.8, is used in current subtraction circuit.
To further reduce the error resulting from variation of VDS, the DC operation voltage
at the output of the current subtraction circuit is set to be 2.5 V. This is achieved by
adjusting the transistors’ aspect ratios of each current mirror to make the DC
operating points of all mirrors to 2.5 V. Furthermore, the non-inverting input of
integrator is tied to 2.5 V to force the circuit operation condition. Although the
mismatch or other process variations may cause the drift of the DC operation voltage
of the current mirrors and therefore the non-zero of Idiff, the drift is much smaller than
the fixed pattern noise of the microbolometer array. It would be corrected with the
on-chip FPN correcting circuit.
Fig. 5.8. Current subtraction circuit.
98
Chapter5. Readout IC for Microbolometer with Current-mode Self-heating Cancellation
Figure 5.9 shows the AC (window 1) and DC sweep (window 2) simulation results of
the current subtraction circuit in Fig. 5.8, where the x axis in DC sweep curve is the
output voltage swept from 2.25 V to 2.75 V. In the simulation, the ramp current
generator is disabled to evaluate the current subtraction circuit independently. The
input is an AC current source plus a DC current offset, which equals to Iref. It is
observed that the DC offset of Idiff is 0.2 nA. The 3 dB bandwidth of the circuit
obtained from the AC sweep analysis of the current subtraction circuit is 3 MHz.
In our design, the voltage variation at the output of the drain (or gate) of M6, which is
mainly caused by the ramp current generator, is generally smaller than 5 mV. Hence,
its effect to the Idiff can be ignored. The current-mode readout circuit is designed to
handle the large offset current of more than ± 35 µA. Thus, the small offset of 0.2 nA
will not affect the dynamic range of circuit and is handled as the noise of FPN. For a
128 × 128 microbolometer FPA, to keep the frame rate higher than 30 frames/sec, the
circuit bandwidth of 500 kHz is required for serial readout. For partial parallel
readout, the requirement should be even lower. The bandwidth of the current
subtraction circuit is enough for the application of the current-mode readout circuits
of microbolometer arrays.
Figure. 5.10 shows the noise analysis results of the current subtraction circuit. The
equivalent input current noise of the circuit is less than HznA8.1 , and the referred
input noise current integrated from 1 Hz to 1 MHz can be calculated as
nAGain
dfI
I Mfoutn
inn 184.15 1
10561.230 18
21
1~1
2_
_ ≈×
=⎟⎟
⎠
⎞
⎜⎜
⎝
⎛
=−
=∫
99
Chapter5. Readout IC for Microbolometer with Current-mode Self-heating Cancellation
Fig.5.9. (a) AC sweep response and (b) DC sweep of the current subtraction circuit.
Fig.5.10. Noise analysis of the current subtraction circuit:
(a) Equivalent input current noise density, (b) output noise power density and (c) total noise power in the bandwidth from 1Hz to 1MHz
100
Chapter5. Readout IC for Microbolometer with Current-mode Self-heating Cancellation
The input current due to the weak infrared signal normally is around 10 nA.
Therefore, the noise of the current-mode readout circuit should be further reduced.
By introducing a current gain in the current mirror (M2 ~ M5) and (M6 ~ M9), the
noise caused by transistors (M10 ~ M13) can be reduced by a factor of the current
gain. Fig. 5.11 shows the noise simulation results of the current subtraction circuit
with a gain factor of 4. The equivalent input current noise now is reduced to less than
HznA56.0 . The input referred current noise integrated from 1 Hz to 1 MHz is
also reduced to
nA
I inn
81.4 4
10036.370 18
_
≈
×=
−
However, the noise reduction is obtained at the cost of the larger chip area consumed
by M3, M5, M7 and M9, whose sizes are 4 times larger than their values in the unit
gain current mirrors. The aspect ratios of other transistors may need to be adjusted to
maintain the desired DC operation condition.
Fig.5.11. (a) Equivalent input current noise density; (b) output current noise power density and (c) total output noise power in the bandwidth from 1 Hz to 1 MHz.
101
Chapter5. Readout IC for Microbolometer with Current-mode Self-heating Cancellation
5.2.3. Current steering DAC design
There are three structures which can be employed for current-steering digital-to-
analog converters (DAC), namely, binary-weighted, thermometer-coded and segment
architecture. For single-ended current output D/A converter, the output current can be
described as
LSBn
nout IbbbI ⋅+++= −− )2 22( 11
11
00 LL (5-7)
Where b0~n-1 are the n-bit input data, ILSB is the current corresponding to the least
significant bit. The output current range is therefore from 0 to
. The design of the DAC will be described in the following
sub-sections.
LSBI⋅+++ )222( L n−110
5.2.3.1. Selection of the current-steering DAC structure
The simplest current-steering DAC structure is the binary-weighted architecture,
which is shown in Fig. 5.12. The input data bits directly control the current mirrors,
which are binary-weighted to their corresponding data bits, through the switching
latch. In this structure, there is no need for the decoder and the number of the latch
and current mirror units are minimized. Thus, the silicon area would be highly
reduced. However, the difference in the size of the current mirror makes the
mismatching a big problem, especially with the increasing of input data bits. That will
cause large INL (Integral Nonlinearity) and DNL (Differential Nonlinearity) errors.
Moreover, severe glitches will arise at the code transitions.
102
Chapter5. Readout IC for Microbolometer with Current-mode Self-heating Cancellation
Fig. 5.12. Binary weighted architecture of the current-steering DAC.
Fig. 5.13. Thermometer coded architecture of the current-steering DAC.
In thermometer-coded architecture, only current sources with the size of one LSB are
used so that the requirement on mismatching can be greatly relaxed and glitches at the
code transitions are minimized. A binary to thermometer decoder is needed to
convert the binary input to 2n-1 control bits. Moreover, the number of the current
sources and latches is increased to 2n-1 for one n-bit D/A converter. Thus, the
complexity of the circuit is increased.
103
Chapter5. Readout IC for Microbolometer with Current-mode Self-heating Cancellation
The trade-off between circuit complexity and mismatching results in the segmented
DAC architecture, where binary-weighted and thermometer coded structures are
combined to obtain an enhanced overall performance. An 8-bit segmented current-
steering DAC is shown in Fig. 5.14. In this example, an 8-bit input is partitioned into
four segments. Every two input bits are converted to three control bits with a simple
binary to thermometer decoder. Four different current sources, representing 1 LSB, 4
LSB, 16 LSB and 64 LSB, respectively, are used. Bit0 and bit1 control three current
sources with the size of 1 LSB; Bit 2 and bit3 control the current sources with size of
4 LSB; and so on. Although thermometer-coded segments are adopted to reduce the
number of current sources having different sizes, the segmented structure still
maintains the advantages of relatively small size and simple circuit structure. Thus
the segmented current steering DAC is chosen for the current-mode readout circuit.
Fig. 5.14. Segmented architecture of the 8-bit current-steering DAC.
5.2.3.2. Current sources and mirrors
104
Chapter5. Readout IC for Microbolometer with Current-mode Self-heating Cancellation
In current-steering DAC, a high-precision current reference independent of the CMOS
process and temperature variation is required. In this design, a bandgap voltage
reference is used to generate the reference current. The schematic of the bandgap
reference circuit is shown in Fig. 5.15[139]. The dimensions of the PMOS transistor
M0, M1 and M2 are same and their gates are connected to a common node. The
operational amplifier is a two-stage amplifier with the PMOS input transistor pair and
Miler compensation. The input current of the opamp can be ignored because of the
zero gate current of the MOS input stage. Thus, current I0, I1 and I2 have the same
value. The operational amplifier forces the voltage VA to be equal to the voltage VB.
PNP transistors T0, T1 and T2 are connected as diodes. T0 and T2 have the same
base-emitter junction size, while the size of T1 is 8 times larger than T0 or T2.
From Fig. 5.15,
⎟⎟⎠
⎞⎜⎜⎝
⎛==
STEBA I
IVVV 0
0 ln (5-8)
011
011 8ln RI
IIVRIVV
STEBB +⎟⎟
⎠
⎞⎜⎜⎝
⎛×
=+= (5-9)
where VT is the thermal voltage and IS is the saturation current of the bipolar device.
Since and , then BA VV = 210 III ==
8ln1
01 TV
RI = (5-10)
The output bandgap reference voltage can be written as
( )( ) 1220
122
122
lnln
ln
RITCIVV
RIII
V
RIVV
TG
ST
EBbg
+−+=
+⎟⎟⎠
⎞⎜⎜⎝
⎛=
+=
γ
(5-11)
where γ and C are constants.
105
Chapter5. Readout IC for Microbolometer with Current-mode Self-heating Cancellation
In order to obtain the zero temperature coefficient of the voltage Vbg, the circuit
parameters should be selected so that the derivative of Vbg with respect to T is held
zero at the circuit operation temperature. Since the collector current of T2 in this
circuit is a PTAT (Proportional To Absolute Temperature) current, the derivative of
Vbg will include TIC ∂∂ . Thus,
( )( ) ( )
qk
RR
qk
TVVV
TV
RR
TTV
RIV
TVV
TIR
TT
TI
IVTCI
TV
TV
TGEB
TTT
GEB
TCTbg
0
102
0
1
02
02
21
2
2
8ln
8ln8ln
ln1lnln
++⋅−−
=
∂+⎟⎟
⎠
⎞⎜⎜⎝
⎛−
∂∂
+−
=
∂∂
+⎟⎟⎠
⎞⎜⎜⎝
⎛∂
−∂+
∂∂
+−∂∂
=∂
∂
γ
γ
γγ
(5-12)
Setting the Eq. (5-12) to zero, the following equation can be obtained.
3
0
1 108.18ln1 −×−≈⎟⎟⎠
⎞⎜⎜⎝
⎛+
RR
qk (5-13)
where k is Boltzmann’s constant (1.38×10-23 J/K); q is electron charge.
The ratio R1/R0 is calculated to be
545.90
1 ≈RR
This ratio has been slightly adjusted, according to the simulation results, to make the
temperature coefficient of the output Vbg to be zero at 27 ºC. After adjustment, the R1
and R0 are chosen to be 5000 Ω and 550 Ω.
The simulation result of the output bandgap voltage, swept from -40 ºC to 100 ºC, is
shown in the Fig. 5.16. The bandgap reference output at 27 ºC is 1.232 V and its
temperature coefficient at room temperature is 0. The maximum variation ∆Vbg from
-40 ºC to 100 ºC is less than 3 mV.
106
Chapter5. Readout IC for Microbolometer with Current-mode Self-heating Cancellation
Fig. 5.15. Schematic of the bandgap voltage reference circuit.
Fig. 5.16. Temperature dependence of the bandgap reference voltage output.
The start-up circuit, which is enclosed in the dash line in the Fig. 5.15, is used to
initialize the circuit when the power is switched on. After the power turned on, the
107
Chapter5. Readout IC for Microbolometer with Current-mode Self-heating Cancellation
voltage VX will become smaller than the threshold voltage of the NMOS transistor and
then the M4 will be kept off. Thus, the start-up circuit will not affect the normal
operation of the bandgap circuit.
The basic current cells of current steering DAC are generated using the above
described bandgap voltage as the reference. The voltage-to-current converter circuit
is shown in the Fig. 5.17. The bandgap voltage is impressed on a resistor and a
NMOS transistor, which is serially connected with the resistor, through an operational
amplifier. M5 and M6 are identical and are working in the saturation region. Since
the bandgap voltage is determined by the silicon bandgap energy, the generation of
the low 1 LSB current will require a very large R0 if M6 is not used. The current,
1LSB_ref, is sensed and mirrored by the cascade current mirror M7 ~ M10. Numbers
of 1 LSB basic current outputs can be generated by similar current mirrors. Because
the bandgap voltage is temperature-insensitive reference, the temperature coefficient
of the 1 LSB output current is also comparatively small. The simulation result of the
voltage-to-current converter is shown in the Fig. 5.18. In the temperature range of -40
ºC to 100 ºC, the output current variation is less than 60 nA. Since the 1 LSB current
is designed to 281.4 nA, the variation is smaller than 21 % of 1 LSB within the
operation temperature range, which is acceptable.
The current sources of 4 LSB, 16 LSB and 64 LSB are obtained in a similar way. A
DC offset current source of -127.5 LSB is also generated and added to output current.
A bipolar current DAC output is therefore obtained.
108
Chapter5. Readout IC for Microbolometer with Current-mode Self-heating Cancellation
Fig. 5.17. Schematic of the voltage-to-current converter.
Fig. 5.18. Simulation result of the 1 LSB basic current generator.
5.2.3.3. Decoder logic
The 2-to-3 lines binary to thermometer decoder can be easily realized by simple
combinational logic circuits. The truth table of the combinational logic is shown in
109
Chapter5. Readout IC for Microbolometer with Current-mode Self-heating Cancellation
Table 5.2. The switches that control the current sources are implemented by using
PMOS transistors. The control signal is active low.
Table 5.2. Truth table of the 2-bit binary-to-thermometer decoder.
Inputs Outputs
Bit 1(s1) Bit 0(s0) Line2 Line1 Line0
0 0 1 1 1
0 1 1 1 0
1 0 1 0 0
1 1 0 0 0
Derived from the truth table, the combinational logic of the decoder can be obtained.
102
11
)10(0
ssLine
sLine
ssLine
•=
=
+=
(5-14)
The schematic of the decoder with latches is shown in Fig. 5.19 and the waveforms of
the decoder outputs are shown in the Fig. 5.20. The maximum propagation delay
from input to Line(i) control signals is less than 0.7 ns. The outputs of the latch, Q
and Qn, control the differential switches for the current sources.
Fig. 5.19. 2-bit binary-to-thermometer decoder circuit and latches.
110
Chapter5. Readout IC for Microbolometer with Current-mode Self-heating Cancellation
Fig. 5.20. Waveforms of the decoder inputs, S0, S1 and outputs Line(0-3).
Fig. 5.21. Schematic of the low-crossing latch.
5.2.3.4. Differential switches and low-crossing latch
In order to reduce the glitches in the current output, the differential switches, which
are made of PMOS transistors, are used for each current source. The low-crossing
latch illustrated in the Fig. 5.21 is employed to generate the control signals for the
111
Chapter5. Readout IC for Microbolometer with Current-mode Self-heating Cancellation
differential switches. The crossing point of the two outputs of the latch can be
adjusted by the ratio of M0 (M1) to M2 (M4). In this design, the crossing point of the
switching signals is set to a value lower than the switching point of the PMOS
transistors so that the differential switches will not be turned off simultaneously. Fig.
5.22 shows the transition of the two output of the low-crossing latch. It can be seen
that the crossing point of the switching control signals is around 1.4 V.
The clock signal is used to synchronize the timing of switching signals. When the
clock signal is low, the outputs are latched, while the clock goes to high, the positive
output, Q, of the latch follows the input signal and the inverting output is obtained at
the negative output. The latch can also operate in the transparent mode, when the
clock is set to high, so that the glitches from the clock signal can be avoid.
Fig. 5.22. Transient analysis results of the low-crossing latch.
The circuit of differential switches is shown in the Fig. 5.24. PMOS transistor pairs
are used. The current is either steered to i_out or i_outn depending on the output of Q
or Qn. The two extra transistors M3 and M4 are added to isolate the switches from
112
Chapter5. Readout IC for Microbolometer with Current-mode Self-heating Cancellation
the output. Therefore, the fluctuations and charge injection due to parasitic
capacitance can be reduced and hence the glitches at the output. The low crossing
point of 1.4 V avoids both PMOS switches, M1 and M2 in the Fig. 5.23, turning off at
the same time.
Fig. 5.23. PMOS differential switches of the current outputs.
5.2.3.5. Current-steering DAC circuit and performance
The overall block diagram of the current-steering DAC circuit is shown in Fig.5.24.
The bandgap voltage generator and the voltage-to-current converter, Cell i_ref,
generate the reference currents of 1 LSB, 4 LSB, 16 LSB and 64 LSB, together with
the constant current output of 127.5 LSB. Each reference current source is mirrored
to three current cells. The digital inputs are first decoded by the 2-3 line binary-to-
thermometer decoder then latched to the output control signals Q(i) and Qn(i). These
signals control the ON or OFF of the differential switches that determine the current
output from each current cell flowing to i_out or i_outn. The output of the DAC can
be expressed as
113
Chapter5. Readout IC for Microbolometer with Current-mode Self-heating Cancellation
LSBLSB IIDDDoutI ×−⋅×><++×><+×><= 5.127)27 2120(_ 710 L (5-15)
where ILSB is designed to be 281.4 nA so that the full range of the DAC output current
can be utilized to cancel the FPN noise.
Since the output i_outn is unused for the microbolometer current-mode readout
circuit, it is connected to a 5 kΩ on-chip resistor to ground to ensure the continuous
current flow during the operation.
The layout of the DAC is plotted in the Fig. 5.26. The power supplies for digital
circuit and the analog circuit are separated in the layout to reduce the crosstalk
between the digital and the analog circuits.
Fig. 5.24. Block diagram of the current-steering DAC.
114
Chapter5. Readout IC for Microbolometer with Current-mode Self-heating Cancellation
Fig.
5.2
5. L
ayou
t of t
he c
urre
nt-s
teer
ing
DA
C.
115
Chapter5. Readout IC for Microbolometer with Current-mode Self-heating Cancellation
The post-layout simulation results of the current-steering DAC, obtained by
simulating the netlist extracted with “all cap” option, are shown in the Fig. 5.26. It
shows the transient analysis results of the current output, i_out, at input digital codes
from 0 to 255. The update frequency of the input code is set to 500 kHz, which is
applicable for the 128 × 128 microbolometer array operating in 30 Hz frame rate.
Fig. 5.26. Transient analysis of the current-steering DAC.
0 50 100 150 200 250-0.004
-0.003
-0.002
-0.001
0.000
0.001
0.002
0.003
0.004
DN
L(LS
B)
Input Code
Fig. 5.27. Simulated DNL of the current-steering DAC.
116
Chapter5. Readout IC for Microbolometer with Current-mode Self-heating Cancellation
0 50 100 150 200 250-0.008
-0.006
-0.004
-0.002
0.000
0.002
0.004
0.006
Gain error=281.49nA/218.40nA =1.00032
INL
(LS
B)
Input Code
Fig. 5.28. Simulated INL of the current-steering DAC.
The differential non-linearity and integrated non-linearity of the output current are
calculated and plotted in the Fig. 5.27 and Fig. 5.28, respectively. It can be seen that
the DNL of the designed DAC is less than ± 0.003 LSB. The INL is less than ±0.006
LSB, where the INL is calculated after the gain error correction of the factor of
1.00032.
5.2.4. Current-mode fixed pattern noise cancellation circuit
The schematic of the current-mode FPN correction circuit is shown in Fig. 5.29. Its
operation is similar with that in the voltage-mode circuits described in chapter 4. In
calibration mode, the DAC output, i_out, is disconnected from the input of the
integrator by the control signal, Calib/Imaging. The input of the integrator, Idiff,
comes from the current subtraction circuit illustrated in Fig. 5.1. The Rbolom in this
117
Chapter5. Readout IC for Microbolometer with Current-mode Self-heating Cancellation
circuit represents the selected microbolometer element in the FPA, where the pixel-
addressing circuit is not shown in the figure. Rdummy is the dummy resistor which has
the same resistance as the room temperature resistance of the microbolometer. Since
the microbolometer array is shielded from the IR incident, and the self-heating noise
is cancelled by the generated ramp current, the integrator integrates the large FPN of
the array. The resulting FPN output is digitized and stored in the off-chip memory.
When in imaging mode, the microbolometer array is sensing the IR signal and the
self-heating cancellation circuit is still activated. The 8-bit FPN data are read back
from the off-chip memory and converted to the bipolar current, IFPN, which is fed to
the input of the integrator, together with Idiff. Thus if the circuit is designed such that
the Idiff of the pixel (x,y) is equal to IFPN of the pixel (x,y), but in the opposite
direction, the fixed pattern noise will be cancelled.
Fig. 5.29 Schematic of the current-mode FPN correction circuit
118
Chapter5. Readout IC for Microbolometer with Current-mode Self-heating Cancellation
The int_ctrl signal, as described in section 4.3.1.1, makes the integration time shorter
than the pixel readout period so that the transient noise from the pixel addressing and
the glitches from the current-steering DAC can be avoided. The gain_ctrl signal sets
the gain of the integrator. In the calibration mode, the gain_ctrl signal should be low
so the large FPN signal is integrated with the gain of int10
tCC +
I diff , where is the
integration time, and will not saturate the integrator. While in imaging mode, high
integrator gain can be used since the integrated current only contains the IR signal and
the small residual fixed pattern noise. The gain of the integrator can be set to a high
value by turning off M2. The circuit is Fig. 5.29 can only provide a coarse
cancellation (8-bits) of the FPN. The fine cancellation (12-bits) is preformed off-chip.
intt
For evaluation of the fixed pattern circuit, a resistor array of 8 × 8 elements with the
resistance values varying within ± 15% of the nominal value, 5000 Ω, is used to
simulate the fixed pattern noise in the microbolometer FPA. The Rdummy is set to 5000
Ω. The ramp current generator in the current subtraction circuit block is disabled in
order to evaluate the FPN correction only. Post-layout simulation results including
effects caused by parasitic capacitors are listed in the Table 5.3, where ∆R is the
resistance difference between the resistors in the array to simulate the microbolometer
and dummy resistor. It is observed from the Table 5.3 that there is a small offset
existing in the readout circuit. It may be caused by the mismatch of the current
mirrors in the current subtraction circuit. It does not affect the 8-bit current-mode
fixed pattern noise correction since the offset is smaller than the 1/3 LSB of the DAC.
The residual FPN noise after correction is shown in the Fig. 5.30. It can be seen that
the Idiff of ±35.9 µA, which is equivalent to the ± 15.3% resistance non-uniformity in
the FPA, can be reduced to less than ± 140 nA, which is smaller than ± 0.5 LSB.
119
Chapter5. Readout IC for Microbolometer with Current-mode Self-heating Cancellation
Table 5.3. Simulation results of the on-chip current-mode FPN correction circuit
∆R (Ω) Idiff (µA) A/D output Idiff (nA) (After correction)
A/D output (After
correction)
765 -35.916 11111111 -25.40 10000000
750 -35.263 11111101 64.45 01111111
650 -30.965 11101110 -46.07 10000000
550 -26.434 11011101 -115.38 10000000
450 -21.871 11001101 -56.89 10000000
350 -17.138 10111100 -107.76 10000000
250 -12.381 10101011 -132.01 10000000
150 -7.519 10011010 -59.54 10000000
50 -2.490 10001000 -98.81 10000000
0 0.970 01111111 -43.40 10000000
-50 2.671 01110110 -3.34 01111111
-150 7.620 01100100 -121.96 10000000
-250 12.511 01010011 -16.61 10000000
-350 17.253 01000010 -60.11 10000000
-450 21.968 00110001 -128.28 10000000
-550 26.516 00100001 -89.57 10000000
-650 31.033 00010001 -72.45 10000000
-750 35.320 00000010 8.68 01111111
-765 35.971 00000000 79.48 01111111
120
Chapter5. Readout IC for Microbolometer with Current-mode Self-heating Cancellation
-800 -600 -400 -200 0 200 400 600 800-200.0nA
-150.0nA
-100.0nA
-50.0nA
0.0A
50.0nA
100.0nA
150.0nA
Res
idua
l cur
rent
afte
r FP
N c
orre
ctio
n
Resistance Difference (Ω)
Fig. 5.30. The residual FPN current after current-mode FPN correction
5.3. Summary
A current-mode readout integrated circuit for microbolometer with the self-heating
cancellation circuit has been described. The self-heating cancellation is achieved
using a ramp current that replicates the self-heating effect. The ramp current is
generated with a PMOS transistor operating in the triode region and a capacitor that
mimics the thermal capacitance of the microbolometer. Simulation results have
shown that with the properly chosen circuit parameters, especially the ratio of I0/C0,
the self-heating effect can be effectively cancelled. An 8-bit on-chip current-steering
DAC is designed for the coarse FPN correction. Post-layout simulation results show
that the 8-bit current-mode FPN correction circuit can reduce the FPN current to less
than ± 0.5 LSB.
121
Chapter 6.______________ _____________________________Experimental Results
Chapter 6. Experimental Results
The operation of the microbolometer detector is significantly affected by the self-
heating effect [132,133,138]. Since the thermal conductance of the microbolometer is
necessarily minimized to achieve high sensitivity, the self-heating typically results in
a large thermal drift of the detector. Studies in chapter 3 show that the self-heating
noise can be much higher than the desired infrared signal. If there is no proper
cancellation, a large portion of the dynamic range of the subsequent readout circuits
would therefore be sacrificed. Thus, an optimized readout circuit is necessary to
remove the noise generated from the self-heating to achieve the high performance.
Based on the proposed self-heating cancellation technique, a prototype ROIC is
fabricated in a Double-poly Triple-metal 0.6-µm CMOS process. The chip is
packaged in a standard PLCC-84 ceramic package. The microphotograph of the
ROIC is illustrated in the Fig. 6.1. Both voltage- and current-mode readout circuits are
integrated onto the same chip. In order to evaluate the readout circuit with either the
microbolometer FPA or a single microbolometer, the simplified voltage- and current-
mode readout circuits, which include the self-heating cancellation but without the
FPN correction, are also designed separately on the chip. The test set-ups for single
bolometer and FPA measurements are described in this chapter. National
Instrument’s PCI-6025E data acquisition card is used for final IR image acquisition.
An image acquisition software written in LabVIEW is developed for testing the
voltage-mode microbolometer FPA readout circuits.
122
Chapter 6.______________ _____________________________Experimental Results
Fig. 6.1. The microphoto of the ROIC.
6.1. Microbolometers used in measurement
The single microbolometer used in the evaluation of self-heating cancellation circuits
is a bulk-micromachined titanium bolometer with the diameter of 50 µm × 50 µm.
The measured room temperature resistance of the single microbolometer is 2.68 kΩ.
Other parameters provided by Institute of Microelectronics, Singapore, are listed in
Table 3.3.
The microbolometer FPA used in testing is a 128x128 2-D array, fabricated in a
standard CMOS process with the micromachining post-process. Its SEM photo is
shown in Fig. 6.2(a). Each microbolometer pixel in the FPA has 50 µm × 50 µm
diameter and room temperature resistance of around 5 kΩ. The picture of the FPA
mounted inside a 68-pin PLCC socket is shown in Fig 6.2(b). The large grey area in
123
Chapter 6.______________ _____________________________Experimental Results
the centre of this chip is the 128 × 128 microbolometer focal plane array. The FPA
consists of Wheatstone bridge readout circuit, pre-amplifier and X-Y scanning
multiplexers. The supply voltage for the on-FPA circuit and bias voltage for the
bridge are both 5 V. A frame synchronization signal is generated by the on-FPA
digital logic circuit. It is a single positive pulse when the last pixel (128, 128) is
selected. And the negative edge of the pulse indicates the position of the signal from
the first pixel.
However, this only available microbolometer FPA for testing has some drawbacks.
The on-chip readout bridge and preamplifier cannot be used for our test because the
pre-amplifier is saturated by the large fixed pattern noise of the FPA and offset
voltage in the bridge. Thus, the output voltage of the microbolometer was taken from
one arm of the bridge and fed to the ROIC, as shown in Fig. 6.2(c). Since the digital
ground for multiplexers in FPA chip cannot be separated from the analog ground to
the bridge, the output of the FPA could be affected by glitches from the digital circuits
in FPA, and consequently reduces the signal-to-noise (S/N) ratio of the readout
circuit. On the other hand, the foundry that fabricated Ti microbolometer arrays is no
longer available for us and a custom process is beyond the available funding
resources. Thus, it is impossible to design and fabricate another FPA. The evaluation
of the ROIC is therefore constrained to these pre-fabricated microbolometer arrays
that are available at the time of the evaluation.
Both the single bolometer and microbolometer FPA used in measurement were
developed by the joint research project [140] and fabricated by Institute of
Microelectronics, Singapore.
124
Chapter 6.______________ _____________________________Experimental Results
(a) (b)
(c)
Fig. 6.2. (a) SEM photo of the microbolometer on FPA; (b) The microphotograph of the microbolometer FPA chip; (c) On-FPA readout bridge and the output of FPA to
our ROIC.
6.2. Test of the voltage-mode self-heating cancellation circuit with
single microbolometer
The single bolometer readout circuit contains only the self-heating cancellation
circuit. The test set-up is shown in Fig. 6.3. The microbolometer, Rbolom, and the
125
Chapter 6.______________ _____________________________Experimental Results
fixed-value resistor, R1, are placed in a vacuum chamber to reduce the bolometer’s
heat dissipation through the air. The air pressure inside the vacuum chamber is kept
at less than 10-4 Torr. A voltage pulse is generated by a HP Function/Arbitrary
Waveform Generator is supplied to the single bolometer and R1 as the electrical bias.
The amplitude of the pulse is 5 Volts high and the duration is 10 µs. However, its
read period is set to 30 ms so that the microbolometer has enough time to cool down
before the next test cycle. The self-heating cancellation circuit is operated under a 5
V power supply. The gain of the preamplifier is approximately unity. The room
temperature resistance of the microbolometer used in testing is 2.68 kΩ.
Fig. 6.3. The test set-up of voltage-mode self-heating cancellation circuit for single bolometer.
The output, Vout, from the preamplifier with different I0/C0 ratios is plotted in Fig. 6.4.
The tuneable feature of the self-heating cancellation circuit can be clearly observed.
Since the current I0 can be tuned off-chip, a good self-heating cancellation can be
achieved. The large spikes, appearing at the beginning and the end of the bias period,
result from the transitions of ф1 and ф2, which does not affect the performance of the
readout circuit if the integration time of the sequential integrator is shorter than the
126
Chapter 6.______________ _____________________________Experimental Results
bias period of 10 µs (9 µs in the testing) and is set in such a way that the spikes could
be avoided. The outputs of the integrator with and without self-heating cancellation
are plotted in the Fig. 6.5. It can be seen that the output would be saturated if no self-
heating cancellation is employed. If the time constant of the integrator is increased to
avoid saturation, the dynamic range of the circuit would be highly degraded.
To further evaluate the effectiveness of the cancellation, the residue error after the
self-heating cancellation is observed, which is shown in Fig. 6.6. The random
measurement noise is due to the use of off-chip bolometer in the measurement and
can be reduced if the bolometer is integrated with the self-heating cancellation circuit.
In most practical readout circuit, an integrator is added after the preamplifier to
improve the signal-to-noise ratio. With such an arrangement, the random noise can be
further reduced. Thus, in the evaluation of the self-heating cancellation, and the
random noise is ignored and a linear fit of the measured data is used. The linear fit
analysis shows that the slope of Vout after self-heating cancellation is about 14 V/s,
which translates to an output voltage drift of 0.126 mV at the end of 9 µs reading
period. This compares to a 12.3 mV drift (based on a slope of about 1367 V/s as
shown in Fig. 6.4) when no self-heating cancellation is applied. Thus, the self-heating
effect is reduced by at least two orders of magnitude with the proposed voltage-mode
self-heating cancellation circuit. The small DC offset of the Vout results from a slightly
unbalance of the bridge readout circuit.
127
Chapter 6.______________ _____________________________Experimental Results
-2.0u 0.0u 2.0u 4.0u 6.0u 8.0u 10.0u 12.0u2.440
2.460
2.480
2.500
2.520
2.540
2.560
Integration Period
Bias Period
Without self-heating cancellation With proper cancellation Over cancellation
Vout
(V)
Time (s)
Fig.6.4. Measured output from preamplifier with different I0/C0 ratios.
-2 .0 u 0 .0 u 2 .0 u 4 .0 u 6 .0 u 8 .0 u 1 0 .0 u 1 2 .0 u-0 .5 0
0 .0 0
0 .5 0
1 .0 0
1 .5 0
2 .0 0
2 .5 0
3 .0 0
3 .5 0
Vin
t_ou
t (V)
T im e (s )
W ith s e lf -h e a t in g c a n c e lla t io n ; P IR = 0 W ith o u t s e lf -h e a tin g c a n c e lla t io n ; P IR = 0
Fig. 6.5. Measured output from integrator with and without self-heating cancellation.
1.0us 2.0us 3.0us 4.0us 5.0us 6.0us 7.0us 8.0us 9.0us2.512V
2.513V
2.514V
2.515V
2.516V
2.517V Vout (after self-heating cancellation) Linear fit of Vout
Vout = 2.51406 +14.04259*t DC offset = 14.06 mV
Vou
t (af
ter p
rope
r sel
f-hea
ting
canc
ella
tion)
Time
Fig. 6.6. Measured residue error after self-heating cancellation.
128
Chapter 6.______________ _____________________________Experimental Results
6.3. Test of the current-mode self-heating cancellation circuit with
single microbolometer
The test set-up for current-mode self-heating cancellation circuit evaluation is shown
in Fig. 6.7. The function generator provides bias voltage to supply to one terminal of
the microbolometer, while its other terminal is connected to the current-mode self-
heating circuits in the ROIC. The pulse voltage bias generates the pulse current
through the microbolometer. The current gain of the current-mode self-heating
cancellation circuit is unity.
Fig. 6.7. Test set-up of the current-mode self-heat cancellation for single bolometer.
The measured result in Fig. 6.8 shows that the self-heating effect (Id) is well
resembled by its replica (Iref - Iramp), which indicates the good performance of the
current-mode self-heating cancellation circuit. Fig. 6.9 shows the output of the Idiff
integrated by the following integrator when IR input power was set to 175nW. It can
be clearly observed that when there is no self-heating cancellation, the output of the
integrator is dominated by the self-heating power and quickly becomes saturated. The
input infrared signal is corrupted.
129
Chapter 6.______________ _____________________________Experimental Results
Following the same method used for the voltage-mode circuit, the residue error in Idiff
after self-heating cancellation and its linear fit are obtained and shown in Fig. 6.10.
The slope of the linear fit is about 8×10-5 A/s. At the end of 9 µs biasing period, Idiff
only reaches 0.72 nA, compared with -0.56 µA (Fig. 6.7) when there is no self-
heating cancellation. The self-heating effect is reduced by more than two orders of
magnitude. The positive slope of the linear fit curve of the residual error indicates the
cancellation circuit was slightly over tuned. There is also a small DC offset in Idiff,
which is caused by the mismatch of the current mirrors in the circuit. The tunable
feature of the current-mode self-heating cancellation circuit is demonstrated in Fig.
6.11.
0.0s 2.0us 4.0us 6.0us 8.0us 10.0us181.0u
181.5u
182.0u
182.5u Id Iref-Iramp
I dand
(Ire
f-Ira
mp)
(A
)
Time
Fig. 6.8. Measured Id and Iref - Iramp.
130
Chapter 6.______________ _____________________________Experimental Results
0.0 2.0u 4.0u 6.0u 8.0u 10.0u
2.4
2.6
2.8
3.0
3.2
3.4
3.6
3.8
4.0
4.2
4.4 Without bias-heating cancellation;
IR input=175nW With proper cancellation;
IR input=175nW
Vin
t_ou
t (V)
Time (s)
Fig. 6.9. Integrator outputs with and without self-heating cancellation, at input IR
power of 175nW.
1.0u 2.0u 3.0u 4.0u 5.0u 6.0u 7.0u 8.0u 9.0u-100.0nA
-50.0nA
0.0A
50.0nA
100.0nA
150.0nA
200.0nA
250.0nA Idiff after self-heating cancellation Linear fit fo Idiff
Idiff = 2.60151E-8 +(7.95521E-5)*t DC offset = 26.0151nA
I diff a
fter p
rope
r sle
f-hea
ting
canc
ella
tion
Time (s)
Fig. 6.10. Measured residue error (Idiff) after self-heating cancellation.
0.0u 2.0u 4.0u 6.0u 8.0u 10.0u-1.0u
-800.0n
-600.0n
-400.0n
-200.0n
0.0
200.0n
400.0n
600.0n
800.0n
1.0u Without self-heating cancellation With proper cancellation over cancellation
I diff (
A)
Time (s)
Fig. 6.11. Tunable feature of the current-mode self-heating cancellation circuit.
131
Chapter 6.______________ _____________________________Experimental Results
6.4. Test of the voltage-mode ROIC with an external microbolometer
FPA
According to the original research plan, a new two-dimensional Ti microbolometer
FPA was to be designed by Institute of Microelectronics (IME) for the ROIC
described in this work and the ROIC was to be designed at National University of
Singapore based on the specifications of the new FPA given by IME. The two parts
will be integrated either in a hybrid or monolithic form and evaluated. However, due
to the change of research plan at IME, the design of the new FPA was not carried out.
Consequently, the readout circuit can only be tested with the existing mirobolometer
FPAs which have different specifications from the originally planed one. As a result,
the voltage-mode ROIC was tested with the 128 × 128 two-dimensional titanium
microbolometer FPA described in section 6.1. The microbolometer FPA used in the
test has a relatively larger fixed pattern noise than originally considered in the ROIC
design. The digital ground on FPA cannot be totally separated from the analog ground.
This has imposed some restrictions on the evaluation and reduces the signal-to-noise
(S/N) ratio of the readout circuit.
6.4.1. Test set-up of the voltage-mode ROIC
The test set-up of the on-chip voltage-mode readout circuit with microbolometer FPA
is illustrated in the Fig. 6.12, which consists of the interface printed circuit board with
the ROIC chip, the microbolometer FPA chip mounted onto the camera body, and
National Instrument PCI-6025E data acquisition (DAQ) card. The regular DC power
supply provides the power for PCB and FPA chip. The computer with the DAQ card
132
Chapter 6.______________ _____________________________Experimental Results
generates the signals, which start and control the operation, and performs the image
data acquisition. The ROIC performs the voltage-mode readout for the
microbolometer FPA with self-heating cancellation and coarse FPN correction.
Fig. 6.12. Test set-up for voltage-mode readout circuit.
During the test, the microbolometer FPA chip is attached to a vacuum dewar, a
vacuum chamber to prevent the loss of sensitivity due to thermal convection through
air. The optical lens of the IR camera head is a germanium lens with focal length of
50 mm, which is used to focus the IR image of the object. A function generator is
used to generate the master clock signal, CLK, for the ROIC, from which other clock
signals are derived. During the readout, each microbolometer is momentarily biased
and its output is sent to ROIC. Finally, the analog image signal is acquired by PCI-
6025E DAQ card and the digitized image is stored on computer. PCI-6025E has 8
differential analog input channels with 12-bit A/D converter resolution and two 12-bit
D/A channels[141]. Its digital I/O unit has bidirectional digital data lines, which can
be programmed by users. To programming the DAQ card, the LabVIEW graphic
133
Chapter 6.______________ _____________________________Experimental Results
program language is used. The data acquisition program for testing the voltage-mode
readout circuit performs the raw image data acquisition and other processing
functions, including fine calibrating, averaging, noise calculation and image display.
6.4.1.1. Hardware interface of the test system
The hardware interfacing between the FPA, ROIC, other circuits on PCB and the
DAQ card are shown in Fig. 6.13. The FPA chip required two DC power supplies,
VDD and Vbias. VDD is the mian power supply for the on-FPA circuits, while Vbias
is for the microbolometer bridge. The main_clk signal from the ROIC is used as the
pixel clock of the FPA. The IR image signal sensed by microbolometer in each pixel
is the input signal, FPA_out (equivalent to VA in Fig. 4.1) to the voltage-mode readout
circuit. Another output signal from the FPA chip is the frame synchronization signal.
When the main_clk starts, the scan begins from an arbitrary pixel of the array
depending on the status of the horizontal and vertical shift registers. However, a
single positive pulse will be generated when the last pixel (128, 128) is selected.
Thus, the negative edge of the frame synchronization pulse indicates the position of
the signal from the first pixel (1, 1). This signal is used to synchronize the operation
of the ROIC and the data acquisition.
134
Chapter 6.______________ _____________________________Experimental Results
Fig. 6.13. The signal flow diagram of the hardware interface of the test system.
The on-chip voltage-mode readout circuit with the self-heating cancellation and fixed
pattern noise correction has been described in the Chapter 4. The ROIC is mounted
on the PCB, with the off-chip SRAMs, RAM’s address generator and some simple
logic circuits. Fig. 6.14 shows the schematic of the digital circuitry on PCB. In
calibration operation, the fixed pattern noise data output from the Dout port of the
ROIC and stored in the off-chip SRAMs. At the mean time, the analog data
acquisition unit on the DAQ card is disabled. In imaging operation, the FPN data are
read from SRAMs and fed back to the Din port of ROIC for the 8-bit coarse FPN
correction. The computer data acquisition is also activated and triggered by the frame
synchronization signal of the FPA.
135
Chapter 6.______________ _____________________________Experimental Results
The DAQ card sends out a Start/Stop signal from its DIO port to control the operation
of the whole system. When this signal is low, all the registers and counters in ROIC
and on PCB are cleared and no signal is read out. When the Start/Stop signal is high,
ROIC is enabled and generates the clocks to drive the FPA and other circuits on PCB.
Meanwhile the signal from FPA is streamed into ROIC and the coarse FPN correction
(calibration mode) or data acquisition of the DAQ card (imaging mode) is carried out.
A Convert* signal is required by the data acquisition unit of the DAQ card as the A/D
clock, whose falling edge indicates the occurring of and A/D conversion.
The low period of the int_clk is the actual readout period of each microbolometer
pixel. In this period, the amplified pixel signal is integrated to minimize the thermal
(Johnson) noise. The integrated pixel signal output is pipelined into an analog input
channel of the DAQ card and converted into digital image data at the Convert* falling
edge. The Convert* signal is designed such that its period is same as the int_clk so
that in every Convert* cycle only one pixel signal streaming out from the ROIC is
sampled and converted to digital data in the DAQ card. However, they are also
designed to have different phase and duty cycle so that the time interval from the
falling edge of Convert* to the rising edge of int_clk is long enough to satisfy the
setup time of the A/D converter. This also minimizes the noise injection due to the
clock transition and the switching noise from the ф1 and ф2 pulses on ROIC’s self-
heating cancellation circuit. The relationship of the main_clk, int_clk and the
Convert* is sketched in the Fig. 6.15.
136
Chapter 6.______________ _____________________________Experimental Results
Fig.
6.1
4. S
chem
atic
of t
he d
igita
l circ
uit o
n th
e PC
B.
137
Chapter 6.______________ _____________________________Experimental Results
Fig. 6.15. Relationship of main_clk, int_clk and Convert*.
The simple logic circuitry on PCB generates the Convert* signal and other control
signals, which are needed by SRAM (IDT71256) and SRAM address generator, such
as WE* for SRAM and the clock of the address counter. The operation of the PCI-
6025E Data Acquisition card is programmed and controlled by the image acquisition
software, which is described in the next section.
6.4.1.2. Image acquisition software
An image acquisition software was written in LabVIEW [142] to control the operation
of the test system.
Normally, image calibration operation was performed before any image is captured.
Covering the optical lens completely, several frames were taken and averaged to
reduce the random noise of the system. These averaged signals were saved and
served as the calibration data for subsequent imaging to provide the one point fine
correction of the fixed pattern noise. During imaging process, the acquired frame was
subtracted from the calibration frame to remove the FPN and then forms the
calibrated image data. Simple noise analysis was performed for both calibration data
and the calibrated images.
138
Chapter 6.______________ _____________________________Experimental Results
The real time imaging interface of the image acquisition software is shown in Fig.
6.16. Noise analysis results are shown in the Data Analysis Panel. Two image display
windows display the acquired raw image without 1-point FPN correction and the
calibrated image after FPN correction respectively.
Fig. 6.17 shows the flow chart of the image acquisition software. The major functions
of the image acquisition software are initializing the data acquisition hardware, real
time image data acquisition, 2-D frame reconstruction, 1-point FPN calibration, noise
measurement, real time image display, save or load calibration file and image file.
Fig. 6.16. User interface of image acquisition
139
Chapter 6.______________ _____________________________Experimental Results
Fig. 6.17. Flow chart of the Image Acquisition LabVIEW programs (a) Real time
imaging operation; (b) Calibration operation.
6.4.2. Test results of the voltage-mode ROIC
The voltage-mode ROIC was tested using the external FPA chip, which was mounted
inside a vacuum chamber of an IR camera head. The vacuum was kept as low as 10-4
140
Chapter 6.______________ _____________________________Experimental Results
Torr. During the test, the bias period of each microbolometer pixel was set to 10 µs to
achieve integration time of about 8.6 µs. The ROIC and microbolometer FPA were
operated under 5 V power supplies. The digital circuitry and analog circuitry on ROIC
use different power supplies to reduce the crosstalk. The room temperature resistance
of the microbolometers in the FPA is around 5 kΩ. The on-chip dummy resistor of
the Wheatstone bridge to balance the micrbolometer resistance was designed to 5 kΩ.
The gain of preamplifier on the ROIC was approximately set to unity when the circuit
is operated in the calibration mode to avoid the saturation due to large FPN.
The calibration process was performed before entering thermal image mode. Several
frames were acquired with the optical lens closed and no FPN correction was applied.
At the same time, the last frame FPN data was saved in the SRAM, which was later
used for on-ROIC FPN correction. With the optical lens was still covered, the system
executed the calibration operation program again. However, at this time, the FPN data
stored in the SRAM was loaded back to the ROIC. The readout circuit on ROIC now
performs self-heating cancellation and on-chip FPN coarse correction. The DAQ card
acquired the analog output of the ROIC, and the FPN data before and after 8-bit on-
chip coarse correction were displayed and saved.
In the measurement, the FPN data should be first acquired while the optical lens is
closed. The FPN data acquired from the first row in the microbolometer FPA is
shown in Fig. 6.18, where the x-axis indicates the position indexes of the pixels in the
first row and y-axis is the output voltage of the integrator in ROIC. Fig. 6.19 depicts
the 2-D intensity graph of the fixed pattern noise before the coarse correction. The x,
y axes indicate the position indexes of the pixels in the corresponding rows and
columns, respectively. The intensity bar shows the output voltage of the integrator,
141
Chapter 6.______________ _____________________________Experimental Results
which represents the amplified FPN of each pixel. It can be observed that there is
considerable variation of signals from different pixels that has been coded into
different blue color scale, even though there is no incoming radiation. Furthermore,
either from the waveform chart of the first raw signals or the 2-D graph, it can be seen
that pixels of the last column have the output of nearly zero. It might be caused by
the bad pixels or the faulty multiplexer circuit in the FPA. Thus, this last column is
excluded in the measurement. The integrator output from ROIC is adjusted to around
0 ~ 5V in order to fully use the circuit dynamic range.
Fig. 6.18. Output waveform chart of the first row of FPA without FPN correction.
Fig. 6.19. Measured 2-D intensity graph of the FPA without coarse FPN correction.
142
Chapter 6.______________ _____________________________Experimental Results
The same measurements are repeated for the case with the 8-bit on-chip coarse FPN
correction applied. The residual FPN outputs of the first raw pixels are shown in Fig.
6.20. It can be seen that the non-uniformity of the FPN outputs are greatly reduced
when the 8-bit on-chip fixed pattern noise correction is applied. The effect of coarse
FPN correction can be clearly observed from the intensity graph acquired in Fig. 6.21
where (a) uses the same intensity scale as that in Fig. 6.19 and (b) shows a zoom-in
version. The first pixel of the frame (at the left-bottom corner) and the pixels in the
last column are excluded in the measurement. The former is due to an error in the
synchronization, while the latter could be due to the defect in the FPA.
Fig. 6.20. The residual FPN readings of the first raw of FPA.
(a) (b)
Fig. 6.21. The residual fixed pattern of the FPA after on-chip correction: (a) Intensity scale is same as that in Fig.6.19 and (b) Intensity scale is reduced.
143
Chapter 6.______________ _____________________________Experimental Results
Fig. 6.22. The histogram of the residual FPN.
The noise measurement program gives the maximum and minimum values of residual
FPN data, which are 2.5934 V and 2.3702 V, respectively. The mean of the residual
FPN is 2.5203 V and the standard deviation (σ) is 0.0342V. The 8-bit on-chip FPN
correction circuit is expected to correct the FPN in range of 0 to 5 V to less than ± 1
LSB, or ± 19.6 mV. However, the residual FPN are within the range of -7 LSB to 5
LSB. The histogram in Fig. 6.22 shows the distribution of the residual FPN.
Although the on-chip FPN correction can only provides 5½ -bit correction, there are
85.6% pixels are corrected to lower than ±2 LSB and 93.2 % pixels to lower than ±3
LSB.
The slightly large residual FPN might result from the poorly fabricated
mcirobolometer FPA which has an unexpectedly large fixed pattern noise measured to
be approximately in range of ± 600 mV at the input of ROIC, while the maximum
input referred FPN the ROIC can handle is ±130 mV.
144
Chapter 6.______________ _____________________________Experimental Results
If those pixels whose original FPN are out of range of ±130 mV are excluded in the
analysis of the residual FPN, better results can be obtained. The maximum and
minimum values of residual FPN data become 2.5573 V and 2.4812 V, respectively.
The standard deviation is now 0.0180 V. Thus, the residual FPN are in range of -1
LSB to 3 LSB. Thus, the correction to 7-bit resolution is achieved.
In order to properly evaluate the performance of the on-chip FPN correction circuit,
an on-chip dummy bolometer array which contains 32-element polysilicon resistors as
the pixels was used in the measurement. The resistance values of these resistors are
within about ±10% variation of 5 kΩ to mimic the typical non-uniformity observed in
microbolometer FPAs. A ±10 % non-uniformity of pixel resistance gives a FPN of
±130 mV in our Wheatstone bridge circuit. The FPN was first digitized and stored in
the off-chip memory during the calibration and later used for the correction. After
correction, the 8-bit digitized residual offset noise at the output of the A/D converter
is shown in the Fig. 6.23.
Fig. 6.23. Residual error of FPN correction when tested with a resistor array.
145
Chapter 6.______________ _____________________________Experimental Results
It can be seen that the on-chip FPN correction reduces the offset noise to less than ±1
LSB. After the 8-bit on-chip correction, the input referred residual error is within ±1
mV. The results prove that the designed 8-bit fixed patter noise correction circuit is
working properly if the condition of the input signal is within the operation limits.
The second step of the measurement is the software calibration of the residual FPN
and the real-time IR image acquisition. With the on-chip FPN correction applied, the
gain of the readout circuit can be increased by reducing the time constant of the on-
chip integrator. While the lens is closed, the calibration data, which is the residual
FPN with the increased gain, are stored in computer for subsequent software fine
calibration in real time imaging operation. When real time imaging is carried out, the
optical lens is open and the calibration data is loaded. The LabVIEW program carries
out the 1-point FPN correction, which is a linear correction by subtracting the stored
residual FPN calibration data from the raw image data to produce a 2-D calibrated
image.
However, since the available microbolometer focal plane arrays are not well matched
with our designed ROIC, the performance of the whole readout system is not
satisfied. The real-time IR image could not be clearly observed. One of the possible
reasons, as mentioned before, is that the FPN of the microbolometer array under test
is too large to be handled by our readout circuit (± 600 mV cf. ± 130 mV). Thus, the
performance of the readout circuit is greatly degraded and the IR signal is
overwhelmed by the large residue FPN. When the overall gain of the ROIC is
reduced to around 400 to accommodate the FPN, the useful IR signal cannot be
adequately amplified. Typical IR signals from room temperature objects are of the
order of several millivolts. With the gain of 400, the 12-bit ADC on the DAQ card is
146
Chapter 6.______________ _____________________________Experimental Results
unable to resolve such a small signal. Furthermore, the noise introduced by the long
wire connections between the FPA in the vacuum dewar and PCB, and by the shared
digital and analog ground in FPA will further reduce the S/N (signal-to-noise) ratio.
These are the main reasons that make the IR signals unreadable.
6.5. Test of the on-chip current-mode FPN correction
Since the available FPA cannot be used to evaluate the current-mode ROIC (due to
the built-in on-chip readout circuit shown in Fig. 6.10), the current-mode readout
circuit is only tested with a built-in 8 × 8 resistor array as a dummy FPA since the
microbolometer FPA for current-mode output is unavailable. The variation of the
resistor values was designed to be within ±10 % of 5 kΩ, which gives a FPN range of
±26.5 µA, similar to the case in the voltage-mode readout circuit.
The test set-up of the current-mode FPN correction circuit is similar to the voltage
mode and shown in Fig. 6.24. The ROIC is mounted on a PCB, which includes logic
circuits, 2 blocks of SRAM and their address generators. The off-chip SRAM1 is used
to store the FPN and calibrated data, while the block SRAM2 is built on PCB to store
the calibrated data after the on-chip FPN correction. A function/arbitrary waveform
generator is used for CLK signal generation. A mixed-signal oscilloscope displays the
ROIC output.
147
Chapter 6.______________ _____________________________Experimental Results
Fig. 6.24. Test set-up for current-mode FPN correction circuit on ROIC.
The first step of the on-chip coarse FPN correction for the current-mode readout
circuits is the calibration operation, while the connection between the output of the
current-steering DAC and the input of the integrator is disconnected. During the
calibration mode, the FPN generated by the resistor array was written into one of the
off-chip SRAMs. During imaging operation, the FPN data were read from SRAM
and fed back to the output of the current subtraction circuit through the on-chip
current DAC. The connection between the DAC and integrator was connected so that
the calibration current output from DAC compensates the offset current, Idiff. The
residual error current was integrated by the integrator and digitized by on-chip ADC
again. The data generated in this operation mode was stored in another off-chip
SRAM. The on-chip ramp current generator was disabled during the test since the
148
Chapter 6.______________ _____________________________Experimental Results
fixed value resistors used to simulate the microbolometer having no self-heating
effect.
The digitized residual FPN after 8-bit on-chip coarse correction is plotted in the Fig.
6.25. It can be seen that the FPN noise is reduced to less than ±2.5 LSB. As the
largest FPN handled by the on-chip feedback circuit is 8-bit, the residual fixed pattern
noise is within ± 0.98% after FPN cancellation, which is near 7-bit resolution. The
results can be made better if the adjustable design for the 1 LSB reference current is
adopted for on-chip current-steering DAC. This may be realized by including a
tunable voltage reference with bandgap source, and convert it to the adjustable
reference current, 1LSB_ref.
The waveforms of the analog output of the current integrator (for several resistor
pixels) before and after FPN correction are shown in the Fig.6.26 (a) and (b),
respectively, where trace D7 is int_ctrl, which controls the integration time of the
current Idiff of each pixel.
0 5 1 0 1 5 2 0 2 5 3 0 3 5 4 0 4 5 5 0 5 5 6 0
-2 L S B
-1 L S B
0 L S B
1 L S B
2 L S B
Res
idua
l FP
N
N o . o f r e s is to r p ix e l
Fig. 6.25. Measured residual error after coarse FPN correction.
149
Chapter 6.______________ _____________________________Experimental Results
(a)
(b)
Fig. 6.26. Output waveform of the on-chip current integrator. (a) before and (b) after coarse FPN correction.
150
Chapter 6.______________ _____________________________Experimental Results
6.6. Summary
Microbolometer ROIC with self-heating cancellation and fixed pattern noise
correction circuits in both voltage- and current-modes were fabricated in a 0.6-µm
standard CMOS process. The fabricated chips were evaluated with both single
bolometer and an external FPA. The effectiveness and tuneable feature of the self-
heating cancellation circuits for both voltage- and current-mode readout have been
demonstrated. When there is no self-heating cancellation, the measured outputs of the
integrator were saturated due to large self-heating noise, which indicates the necessity
of the self-heating cancellation circuits in ROIC.
The voltage-mode ROIC has been measured using the external 128 × 128
microbolometer FPA. The measurement results for the 8-bit on-chip fixed pattern
noise correction shows that it is able to provide 5½-bit correction when tested with the
microbolometer FPA. The degradation in the performance of the correction with FPA
is believed to be caused by the unexpected large non-uniformity of the
microbolometer FPA. If the pixels with large FPN were excluded in measurements,
the 7-bit resolution can be achieved. The FPN correction circuit attains 8-bit
correction when measured with an on-chip resistor array.
Since the existing microbolometer FPA is not suitable for the current mode ROIC, the
current-mode FPN correction circuit has been measured with an on-chip resistor
array. The results show that the on-chip coarse FPN correction can reduce the FPN to
less than ±2.5 LSB.
151
Chapter 6.______________ _____________________________Experimental Results
Since the integration of microbolometer FPA and ROIC could not be carried out, the
large FPN and readout circuit in the existing microbolometer FPA have, in some way,
prevented the fabricated voltage- and current-mode ROIC from being fully evaluated.
The large FPN is believed to be the main reason for which the real-time IR image
could not be detected.
152
Chapter 7.__________________________________________________Conclusions
Chapter 7. Conclusions
7.1. Conclusions
A new self-heating cancellation technique suitable for microbolometer IR detectors
and arrays has been proposed in this thesis. The cancellation is based on the
equivalence between the electrical and thermal system, by which the thermal
capacitance can be represented by a capacitance in the electrical domain. Therefore
the self-heating effect in the thermal domain can be easily generated using a current
source and capacitor, which can be later used to cancel the self-heating effect itself.
The newly proposed self-heating cancellation concept has been successfully
implemented and demonstrated in a voltage- and a current-mode readout circuit for
single microbolometer. The test results of the fabricated chips have shown that the
self-heating effect can be reduced by at least two orders of magnitude in both voltage-
and current-mode readout circuits.
A SPICE electro-thermal model of microbolometer has been constructed to study the
behaviours of the microbolometer and gain better understanding of the self-heating
effect. The model has been validated by the experiment data and used in the design of
the self-heating cancellation circuits.
The ROIC with the proposed voltage-mode self-heating cancellation circuit has been
developed together with an 8-bit FPN correction circuit. The ROIC is fabricated in a
0.6 µm CMOS technology and evaluated with an external 128 × 128 microbolometer
153
Chapter 7.__________________________________________________Conclusions
FPA. The test results have shown that the FPN can be corrected to a 5½-bit resolution
against the expected design specification of 8 bits. This is attributed to the
unexpected large FPN in the IR FPA under test, which is observed during the test.
When this factor is excluded from the test, the resolution of the correction is improved
to 7 bits. This has also been approved by a test in which the FPA is replaced by a
resistor array, whose resistance variation resembles the FPN in the microbolometer
FPA. The correction of 8-bit has attained in this case.
The current-mode self-heating cancellation circuit has also been incorporated in a
current-mode ROIC with an 8-bit FPN correction. However, since the microbolometer
FPA for current-mode ROIC is not available, the fabricated ROIC can only been
evaluated with a built-in 8 × 8 resistor array for the performance of FPN correction.
The measured residual FPN of less than ±2.5 LSB has been achieved.
7.2. Original contributions
The concept of using the equivalence between the electrical and thermal system for
the cancellation of the self-heating in microbolometer is first proposed in this thesis.
The voltage- and current-mode self-heating cancellation circuits have been
successfully demonstrated and the results are published in [136] and [138]. A
Singapore and US Provisional patent have been filed for the voltage- and current-
mode self-heating cancellation circuits, respectively. In addition, the proposed
concept can be applied to the design of the readout circuits for other types of sensors,
where similar problem exists.
7.3. Future work
154
Chapter 7.__________________________________________________Conclusions
The evaluation of the ROIC in this thesis is restricted by the availability of the
suitable microbolometer FPAs. Thus, to properly evaluate the ROIC with the
proposed self-hearing cancellation circuit, the integration of the readout circuit with a
microbolometer focal plane array should be further pursued. To reduce the FPN, an
option to place the load resistor (in Wheatstone bridge) in the pixel and near the
detector using same material as the microbolometer, can be explored. With such an
arrangement, the non-uniformity or FPN of the FPA is expected to be greatly reduced.
This is because the resistance ratio of the bolometer and its load can be kept at
relatively constant. Moreover, the low noise design approach should be adopted in
the future implementation of the ROIC to ensure the high signal-to-noise ratio
required for IR imaging at room temperature.
155
References____________________________________________________________
References
[1] Schlessinger, M. Infrared Technology Fundamentals. – 2nd edition. pp.1-25,
Marcel Dekker Inc. 1995.
[2] Burnay, S.G., T.L. Williams and C.H. Jones. Applications of Thermal Imaging.
Adam Hilger, Bristol. 1988.
[3] Rogalski, A. and K. Chrzanowski. Infrared Devices and Techniques, Opto-
Electronics Review, Vol.10(2), pp.111–136. 2002.
[4] Dereniak, E. L. and G. D. Boreman. Infrared Detectors and Systems. pp.395-
438, John Wiley and Sons, 1996.
[5] P.R. Norton, Infrared Detectors in the next millennium, Proc. SPIE, Vol.3698,
pp.652–665. 1999.
[6] Kaplan, H. Practical Applications of Infrared Thermal Sensing and Imaging
Equipment, SPIE Optical Engineering Press, 1993.
[7] Dempsey, R.J., D.G. Davis, R.G. Jr. Bruice and R.A. Lodder. Biological and
Medical Applications of Near Infrared Spectrometry, Applied Spectroscopy,
Vol.50(2), pp.18A-34A, 1996.
[8] Kruse, P.W. Uncooled Thermal Imaging Arrays, Systems, and Applications,
SPIE Press, 2001.
[9] BAE Systems Information and Electronic Systems Integration Inc., Uncooled
imagery, Mar. 2003. URL: http://www.iews.na.baesystems.com/iris/unc-
imagery.htm. (Accessed on Apr. 22nd, 2003).
156
References____________________________________________________________
[10] He, X., G. Karunasiri, T. Mei, W. J. Zheng, P. Neuzil and U. Sridhar.
Performance of Microbolometer Focal Plane Arrays under Varying Pressure,
IEEE Electron Device Letters, Vol.21, pp.233-235. 2000.
[11] McDaniel, G. W. and D. Z. Robinson. Thermal Imaging by Means of the
Evaporagraph, Appl. Opt., Vol.1, pp.311-324. 1962.
[12] Burnay, S. G., T. L. Williams, and C. H. Jones. Applications of Thermal
Imaging, Adam Hilger, Bristol. 1988.
[13] Kruse, P. W. Thermal Imagers Move from Military to Marketplace, Photonics
Spectra, Vol. 29, No. 3, p.103. 1995.
[14] Liddiard, K. C. Thin-film Resistance Bolometer IR Detectors, Infrared Physics,
Vol.24, pp.57-64. 1984.
[15] Pommerenig, D. H., D. D. Enders and T. E. Meinhardt. Hybrid Silicon Focal
Plane Array Development: An Update, Proc. Society Pho-Opt. Instrumentation
Engineering, Vol.267, pp.23–30. 1981.
[16] Rogalski, A. Infrared Detectors. pp.3-152, Gordon and Breach Science
Publishers, Amsterdam. 2000.
[17] Rogalski, A. Infrared Detectors: an Overview, Infrared Physics & Technology,
Vol.43, pp.187 –210. 2002.
[18] Kinch, M. A., S. R. Borrello, D. H. Breazeale and A. Simmons. Geometrical
Enhancement of HgCdTe Photoconductive Detectors, Infrared Phys., Vol.17,
pp.137-145. 1977.
[19] Musca, C. A., J. F. Siliquini, K. A. Fynn, B. D. Nener, L. Faraone and S. J. C.
Irvine. MOCVD-grown Wider-bandgap Capping Layers in Hg1-xCdxTe Long-
157
References____________________________________________________________
wavelength Infrared Photonconductors, Semicon Sci. Technolo. Vol.11,
pp.1921-1922. 1996.
[20] Smith, D. L., D. K. Arch, R. A. Wood and M. W. Scott. HgCdTe
Heterojunction Contact Photoconductor, J. Appl. phys., Vol. 45, pp. 83-85.
1984.
[21] Gunapala, S. D., S. V. Bandara, J. K. Liu, E. M. Luong, S. B. Rafol, J. M.
Mumolo, D. Z. Ting, J. J. Bock, M. E. Ressler, M. Levy, N. LeVan, M. A.
Fauci. Recent Developments and Applications of Quantum Well Infrared
Photodetector Focal Plane Arrays, Opto-Electron. Rev., Vol. 9, pp.150 –163.
2001.
[22] Andersson, J. Y., Dark Current Mechanisms and Conditions of Background
Radiation Limitation of N-doped AlGaAs/GaAs Quantum-well Infrared
Detectors, J. Appl. Phys. Vol.78, pp. 6298–6304. 1995.
[23] Piotrowski, J., W. Galus and M. Grudzien. Near Room-temperature
Photodetectors, Infrared Phys. Vol.31, pp.1-48. 1991.
[24] Butler, N., R. Blackwell, R. Murphy, R. Silva and C. Marshall. Low Cost
Uncooled Microbolometer Imaging System for Dual Use, Proc. SPIE, Vol.
2552, pp. 583–591. July 1995.
[25] Foote, M. C., E. W. Jones, T. Caillat. Uncooled Thermopile Infrared Detector
Linear Arrays with Detectivity Greater than 109 cmHz1/2/W, IEEE Trans.
Electron Devices, Vol. ED-45, pp.1896-1902. 1998.
[26] Almasri, M., D. P. Butler, and Z. Çelik-Butler. Self-supporting Infrared
Microbolometers with Low Thermal Mass, IEEE/ASME Journal of
Microelectromechanical Systems, Vol. 10, pp. 469-476. 2001.
158
References____________________________________________________________
[27] Baer, W. G., K. Najafi, K. D. Wise and R. S. Toth. A 32-element
Micromachined Thermal Imager with On-chip Multiplexing, Sensors &
Actuators A, Vol. A48, pp.47-54. 1995.
[28] Lenggenhager, R., H. Baltes and T. Elbel. Thermoelectric Infrared Sensors in
CMOS Technology, Sensors & Actuators A, Vol A37-38, pp.216-220. 1993.
[29] Parameswaran, M., H. Baltes, L. Ristic, D. Dhaded and A. Robinson. A New
Approach for the fabrication of Micromechanical Structures, Sensors &
Actuators A, Vol. 19, pp.289–307. 1989.
[30] Howard, P. E., C. J. Han, J. E. Clarke, J. C. Stevens, P. A. Ely and E. T.
Fitzgibbo. Advances in Microbolometer Focal Plane Technology at Boeing,
Proc. SPIE, Vol. 3379, pp.47-57. 1998.
[31] Howard, P. E., J. E. Clarke, W. J. Parrish and J. T. Woolaway. Advanced High-
Performance 320×249 VOx Microbolometer Uncooled IR Focal Plane Array,
Proc. SPIE, Vol. 3698, pp.131-136. 1999.
[32] Radford, W., D. F. Murphy, J. A. Finch, K. Hay, A. Kennedy, M. Ray, A. A.
Sayed, J. Wyles, R. Wyles, J. B. Varesi, E. A. Moody and F. Cheung.
Sensitivity Improvements in Uncooled Microbolometer FPAs, Proc. of SPIE,
Vol.3698, pp.119-130. 1999.
[33] Breen, T. B., N. R. Butler, M. Kohin, C. A. Marshall, R. Murphy, T. W. Parker
and R. J. Silva. Applications of Uncooled Microbolometer Sensors, Proc. SPIE,
Vol. 3379, pp.145-156. 1998.
159
References____________________________________________________________
[34] Kohli, M., C. Wuthrich, K. G. Brooks, B. Willing, M. Forster, P. Muralt, N.
Setter and P. Ryser, Pyroelectric Thin Film Sensor Array, Sensors &
Actuators A, Vol.60, pp.147–53. 1997.
[35] Hanson, C. M. Uncooled Thermal Imaging at Texas Instruments, Proc. SPIE,
Vol.2020, pp.330 –339. 1993.
[36] Noda, M., T. Miyamoto, S. Murakami, T. Nomura, K. Inoue and M. Okuyama.
A Dielectric Bolometer Mode of Infrared Sensor Using a New BaTi1-xSnxO3
Thin Film with a High Temperature Coefficient of Dielectric Constant,
Integrated Ferroelectrics, Vol.49, pp.305-314. 2002.
[37] Hashimoto, K., H-P. Xu, T. Mukaigawa, R. Kubo, H. Zhu, M. Noda and M.
Okuyama. Si Monolithic Microbolometers of Ferroelectric BST Thin Film
Combined with Readout FET for Uncooled Infrared Image Sensor, Sensors &
Actuators A, Vol.88, No.1, pp.10-19, Jan. 2001.
[38] Anthony, C. J., M. A. Todd, P. P. Donohue, M. A. C. Harper, Z. Huang and J.
C. Jones. High Performance PST Thin Films on Polysilicon Sacrificial Layers
for Uncooled IR Detectors, Integrated Ferroelectrics, Vol. 49, pp.233-243. 2002.
[39] Huang, Z., P. P. Donohue, Q. Zhang, D. J. Williams, C. J. Anthony, R. W.
Whatmore and M. A. Todd. Comparative Microstructure and Electrical
Property Studies of Lead Scandium Tantalite Thin Films as Prepared by
LDCVD, Sol-gel and Sputtering Techniques, J. Phys. D – Appl. Phys., Vol.36,
pp.270-279. 2003.
[40] Kim, D. H. and H. S. Kwok. Pulsed Laser Deposition of VO2 Thin Films,
Appl. Phys. Lett., Vol.65, pp.3188-3190. 1994.
160
References____________________________________________________________
[41] Chen, C.-H., X.-J. Yi, X.-R. Zhao and B.-F. Xiong. Characterizations of VO2-
based Uncooled Microbolometer Linear Array, Sensors & Actuators A, Vol. 90,
pp.212-214. May 2001.
[42] Jerominek, H., F. Picard and D. Vincent. Vanadium Dioxide Films for Optical
Switching and Detection, Opt. Eng., Vol.32, pp.2092-2099. 1993.
[43] Kuzma, E. Contribution to the Technology of Critical Temperature Resisters,
Electron Technology, Vol. 26, pp.129-142. 1993.
[44] Wood, R.A. Micromachined Bolometer Array Achieve Low-cost Imaging,
Laser Focus World, Vol. 29, pp.101-106. June, 1993.
[45] Wood, R.A. Uncooled Thermal Imaging with Monolithic Silicon Focal Planes,
Proc. SPIE, Vol. 2020, pp.322-329. 1993.
[46] Jerominek, H., F. Picard, N. R. Swart, M. Renaud, M. Levesque, M. Lehoux,
J.-S. Castonguay, M. Pelletier, G. Bilodeau, D. Audet, T. D. Pope and P.
Lambert. Micromachined, Uncooled, VO2-based IR Bolometer Arrays, Proc.
SPIE, Vol. 2746, pp.60-71. 1996.
[47] Jerominek, H., T. D. Pope, C. Alain, R. Zhang, F. Picard, M. Lehoux, F. Cayer,
S. Svard, C. Larouche and C. Grenier, Miniature VO2-based Bolometric
Detectors for High-resolution Uncooled FPAs, Proc. SPIE, Vol. 4028, pp. 47-
56. 2000.
[48] Tanaka, A., S. Matsumoto, N. Tsukamoto, S. Itoh, K. Chiba, T. Endoh, A.
Nakazato, K. Okuyama, Y. Kumazawa, M. Hijikawa, H. Gotoh, T. Tanaka and
N. Teranishi. Infrared Focal Plane Array Incorporationg Silicon IC Process
Compatible Bolometer, IEEE Trans. Electron Devices, Vol.43, pp.1844-1850.
Nov. 1996.
161
References____________________________________________________________
[49] Lambkin, P., B. Lane, I. O'Heifearnan, J. Gillham and R. Watton.
Characterization of CMOS Compatible Uncooled Microbolometers, In Proc.
IEEE/LEOS Int. Conf. on Optical MEMS, Aug. 2000, Kauai, USA, pp.99-100.
[50] Ju, S-B., Y-J. Yong and S-G. Kim. Design and Fabrication of a High-fill-factor
Microbolometer Using Double Sacrificial Layers, Proc. SPIE, Vol. 3698,
pp.180-189. 1999.
[51] Lee, H.-K., J.-B. Yoon, E. Yoon, S.-B. Ju, Y.-J. Yong, W. Lee and S.-G. Kim.
A High Fill−factor Infrared Bolometer Using Micromachined Multilevel
Electrothermal Structures, IEEE Trans. Electron. Devices, Vol.46, pp.1489–
1491. 1999.
[52] Liddiard, K. C., U. Ringh, C. Jansson and O. Reinbold. Progress of Swedish–
Australian Research Collaboration on Uncooled Smart IR Sensors, Proc. SPIE,
Vol. 3436, pp. 578-584, 1998.
[53] Malyarov, V. G., I. A. Khrebtov, Y. V. Kulikov, I. I. Shaganov, V. Y. Zerov, N.
A. Feoktistov. Comparative Investigation of Bolometric Properties of Thin-
film Structures Based on Vanadium Dioxide and Amorphous Hydrated Silicon,
Proc. SPIE, Vol. 3819, pp.136-142. June 1999.
[54] Tissot, J.-L., F. Rothan, C. Vedel, M. Vilain and J.-J. Yon. LETI/LIR’s
Amorphous Silicon Uncooled Microbolometer Development, Proc. SPIE,
Vol.3379, pp.139-144. 1998.
[55] Ichihara, T., Y. Watabe, Y. Honda and K. Aizawa. A High Performance
Amorphous Si1-xCx:H Thermistor Bolometer Based on Micro-machined
Structure, In Proc. Int. Conf. on Solid-State Sensors & Actuators
(Transducers’97), pp.1253-1256. June 1997.
162
References____________________________________________________________
[56] Mottin, E., J.-L. Martin, J.-L. Ouvried-Buffet, M. Vilain, J.-J. Yon and J.-P.
Chatard. 320×240 Microbolometer Uncooled IRFPA, Proc. SPIE, Vol.4028,
pp.40-46. 2000.
[57] Onokhov, A. P., M. A. Fedorov and N. A. Feoktistov. Photometric Study of
Layers of Amorphous Silicon Carbide, Journal of Optical Technology, Vol. 69
pp.766-768. Oct. 2002.
[58] Rice, J. P., E. N. Grossman and D. A. Rudman. Antenna-coupled High-Tc Air-
bridge Microbolometer on Silicon, Appl. Phys. Lett. Vol.65, pp.773-775. 1994.
[59] Brasunas, J. C. and B. Lakew. High-Tc Superconductor Bolometer with
Record Performance, Appl. Phys. Lett. Vol.64, pp.777-779. 1994.
[60] Gray, J., Z. Celik-Butler, D. P. Butler and M. Almasri. Semiconducting
YBaCuO as Infrared Detecting Bolometers, Proc. SPIE, Vol.3436, pp.555-565.
1998.
[61] Jahanzeb, A., C. M. Travers, Z. Celik-Butler, D. P. Butler, S. G. Tan. A
Semiconductor YBaCuO Microbolometer for Room-temperature IR Imaging,
IEEE Trans. Electron. Devices, Vol.44(10), pp.1795-1801. 1997.
[62] Almasri, M., D. P. Butler and Z. Celik-Butler. Semiconducting YBCO
Bolometers for Uncooled IR Detection, Proc. SPIE, Vol.4028, pp.17-26. 2000.
[63] De Moor, P., J. John, S. Sedky and C. Van Hoof. Linear Arrays of Fast
Uncooled Poly SiGe Microbolometers for IR Detection, Proc. SPIE, Vol.4028,
pp.27-35. 2000.
[64] Sedky, S., P. Fiorini, K. Baert, L. Hermans and R. Mertens. Characterization
and Optimization of Infrared Poly SiGe Bolometers, IEEE Trans. Electron.
Devices, Vol.46, pp. 675–682. 1999.
163
References____________________________________________________________
[65] De Moor, P., S. Sedky, D. Sabuncuoglu and C. Van Hoof. Linear Arrays of
Uncooled Poly SiGe Microbolometers for IR Detection, Proc. SPIE, Vol.3876,
pp.256-259. 1999.
[66] Ishikawa, T., M. Ueno, K. Endo, Y. Nakaki, H. Hata, T. Sone, M. Kimata and
T. Ozeki. Low-cost 320×240 Uncooled IRFPA Using Conventional Silicon IC
Process, Proc. SPIE, Vol. 3698, pp.556-564. 1999.
[67] Kim, J. and C. Han. A New Uncooled Thermal Infrared Detector Using
Silicon Diode, Sensors & Actuators A, Vol.89, pp. 22–27. 2001.
[68] Brasunas, J. C., B. Lakew and R. Fettig. A Comment on Reported Detectivity
of a New Uncooled Thermal Infrared Detector, Sensors & Actuators A, Vol.96,
pp.211–213. 2002.
[69] Tezcan, D. S., F. Koçer and T. Akin. An Uncooled Microbolometer Infrared
Detector in Any Standard CMOS Technology, In Proc. 10th Int. Conf. on Solid-
State Sensors & Actuators (TRANSDUCERS’99), June, 1999. Sendai, Japan,
pp.610-613.
[70] Eminoglu, S., D. S. Tezcan, M. Y. Tanrikulu, T. Akin. Low-cost Uncooled
Infrared Detectors in CMOS Process, Sensors & Actuators A, Vol.109, pp.102–
113. 2003.
[71] Hsieh, C.-C., C.-Y. Wu, F.-W. Jih, and T.-P. Sun. Focal-Plane-Arrays and
CMOS Readout Techniques of Infrared Imaging Systems, IEEE Transactions
on Circuits and Systems for Video Technology, Vol.7, No. 4, pp.594-605 1997.
[72] Balcerak, R. and D. P. Jenkins. Uncooled Infrared Focal Plane Arrays, In Proc.
18th Annual International Conference of the IEEE Engineering in Medical and
Biology Society, 1996, Amsterdam, pp.2077-2078.
164
References____________________________________________________________
[73] Nelson, M. D., J. F. Johnson and T. S. Lomheim. General Noise Processes in
Hybrid Infrared Focal Plane Arrays, Opt. Eng., Vol.30, No.11, pp.1682–1700.
1991.
[74] Agnese, P., C. Buzzi, P. Rey, L. Rodriguez and J.-L. Tissot. New
Technological Development for Far-infrared Bolometer Arrays, Proc. SPIE,
Vol.3698, pp.284-290. 1999.
[75] Kavadias, S., P. De Moor, M. Gastal, N. Hendrikx, F. Wirix and C. A. Van
Hoof. Hybrid Long-wavelength IR Sensor Based on a Linear Array of Poly
SiGe Uncooled Microbolometers with a CMOS Readout, Proc. SPIE, Vol.
4169, pp.348-355. 2001.
[76] Liu, C. C. and C. H. Mastrangelo. A CMOS Uncooled Heat-Balancing Infrared
Imager, IEEE J. Solid-State Circuits, Vol.35(4), pp. 527-535. Apr. 2000.
[77] Zerov, V. Y., Y. V. Kulikov, V. G. Malyarov, I. A. Khrebtov, I. I. Shaganov
and E. B. Shadrin. Vanadium Oxide Films with Improved Characteristics for
IR Microbolometer Devices, Tech. Phys. Lett. Vol.27, pp.378-380. 2001.
[78] Infrared Solutions Inc. IR-160 Thermal Imager, 2004, URL:
http://www.infraredsolutions.com/html/products/ir160F.shtml. (Last accessed
at Feb. 25th, 2004)
[79] Ishikawa, T., M. Ueno, K. Endo, Y. Nakaki, H. Hata, T. Sone and M. Kimata.
Low-cost 320×240 Uncooled IRFPA Using Conventional Silicon IC Process,
Opto-Electron. Rev. Vol.7, pp. 297 –303. 1999.
[80] Radford, W., D. Murphy, M. Ray, S. Propst, A. Kennedy, J. Kojiro, J.
Woolaway, K. Sch, R. Coda, G. Lung, E. Moody, D. Gleichman and S. Baur.
165
References____________________________________________________________
320×240 Silicon Microbolometer Uncooled IRFPAs with On-chip Offset
Correction, Proc. SPIE, Vol.2746, pp.82-92. 1996.
[81] Oda, N., Y. Tanaka, T. Sasaki, A. Ajisawa, A. Kawahara and S. Kurashina.
Performance of 320 × 240 Bolometer-type Uncooled Infrared Detector, NEC
Research & Developement, Vol. 44(2), pp.170-174. 2003.
[82] Radford, W., D. Murphy, A. Finch, A. Kennedy, J. K. Kojiro, M. Ray, R.
Wyles, R. Coda, E. A. Moody and S. T. Baur. Microbolometer Uncooled
Infrared Camera with 20-mK NETD, Proc. SPIE. Vol.3379, pp22-35, 1998.
[83] Wood, R.A., C. J. Han and P. W. Kruse. Integrated Uncooled Infrared Detector
Imaging Arrays, In Technical Digest. IEEE 5th Solid-State Sensor and Actuator
Workshop, 22-25 Jun 1992, USA, pp.132 -135.
[84] Radford, W. A., R. Wyles, J. Wyles, J. B. Varesi, M. Ray, D. F. Murphy, A.
Kennedy, A. Finch, E. A. Moody, F. Cheung, R. Coda and S. T. Baur.
Microbolometer Uncooled Infrared Camera with 20-mK NETD, Proc. SPIE,
Vol. 3436, pp.636-646. 1998.
[85] Marshall, C. A., N. R. Butler, R. Blackwell, R. Murphy and T. Breen.
Uncooled Infrared Sensors with Digital Focal Plane Array, Proc. SPIE, Vol.
2746, pp.23-31. 1996.
[86] Radford, W. A., D. Murphy, M. Ray, S. H. Propst, A. Kennedy, Kevin L. Soch,
R. Coda, G. Lung, E. A. Moody and D. Gleichman. 320×240 Microbolometer
Focal Plane Array for Uncooled Applications, Proc. SPIE, Vol. 2685, pp.80-90.
1996.
166
References____________________________________________________________
[87] Meyer, B., R. F. Cannata, A. Stout, A. Gin, P. Taylor, E. Woodbury, J. Deffner
and F. Ennerson. Amber's Uncooled Microbolometer LWIR Camera, Proc.
SPIE, Vol.2746, pp.13-22. 1996.
[88] Murphy, R., M. Kohin, B. S. Backer, N. R. Butler, R. Blackwell and T. Allen.
Recent Developments in Uncooled IR Technology, Proc. SPIE, Vol. 4028,
pp.12-16. 2000.
[89] Murphy, D. F., M. Ray, R. Wyles, J. F. Asbrock, N. A. Lum, J. Wyles,C.
Hewitt, A. Kennedy, D. Van Lue, J. S. Anderson, D. Bradley, R. Chin and T.
Kostrzewa. High-sensitivity 25-micron Microbolometer FPAs, Proc. SPIE, Vol.
4721, pp.99-110. 2002.
[90] Harling, G., R. Zhang, T. Pope, F. Picard and A. Azelman. Non-contacting
Temperature Sensing Device, U.S. Patent, No. 6222454. 2001.
[91] Ueno, M., O. Kaneda, T. Ishikawa, K. Yamada, A. Yamada, M. Kimata and M.
Nunoshita. Monolithic Uncooled Infrared Image Sensor with 160 by 120
Pixels, Proc. SPIE, Vol. 2552, pp.636-643. 1995.
[92] Parrish, W. J., J. T. Woolaway, G. Kincaid, J-L. Heath and J. D. Frank. Low
Cost 160×128 Uncooled Infrared Sensor Array, Proc. SPIE, Vol. 3360, pp.111-
119. 1998.
[93] Jansson, C., U. Ringh, K. C. Liddiard and N. Robinson. FOA/DSTO Uncooled
IRFPA Development, Proc. SPIE, Vol. 3698, pp.264-275. 1999.
[94] Earth Observatory, NASA, On the Shoulders of Giants, Samuel Langley, URL:
http://earthobservatory.nasa.gov/Library/Giants/Langley/langley_2.html. (Last
accessed on June 18th, 2004)
167
References____________________________________________________________
[95] X. Gu, G. Karunasiri and G. Chen, Determination of the Thermal Parameters of
Microbolometers Using a single Electrical measurement, Applied Physics
letters, Vol.72, No.15, pp.1881-1883. 1998.
[96] Rieke, F.M., A. E. Lange, J. W. Beeman and E. E. Haller. An AC Bridge
Readout for Bolometric Detectors, IEEE Transactions on Nuclear Science,
Vol.36(1), pp.946-949. Feb. 1989.
[97] Wilbanks, T. Improved Low Frequency Stability of Bolometric Detectors,
IEEE Trans. Nuclear Sci., Vol. 37(2), pp.566-572. 1990.
[98] Gaertner, S., A. Benoit, J. M. Lamarre, M. Giard, J. L. Bret, J. P. Chabaud, F.
X. Desert, J. P. Faure, G. Jegoudez, J. Lande, J. Leblanc, J. P. Lepeltier, J.
Narbonne, M. Piat, R. Pons, G. Serra and G. Simiand. A New Readout System
for Bolometers with Improved Low Frequency Stability, Astronomy &
Astrophysics Supplement Series, Vol.126(1), pp.151-160. Nov. 1997.
[99] Yu, T.-H., C.-Y. Wu, P.-Y. Chen, F.-W. Chi, J.-J. Luo, C. D. Chiang and Y.-T.
Cheng. A New CMOS Readout Circuit for Uncooled Bolometric Infrared
Focal Plane Arrays, In Proc. 2000 IEEE International Symposium on Circuits
and Systems(ISCAS 2000), 2000, Geneva, Vol.2, pp.493-496.
[100] Kavadias, S., P. De Moor and C. Van Hoof. CMOS Circuit for Readout of
Microbolometer Arrays, Electronics Letters, Vol.37, No.8, pp.481-482. Apr.
2001.
[101] Chin, Y. L., T. P. Sun, J. C. Chou, Y.-C. Chin, Y. C. Chou, W. Y. Chung and S.
K. Hsiung. High-Performance Readout Integrated Circuit for Surface-
Micromachined Bolometer Arrays, Proc. SPIE, Vol. 3899, pp.124-132.1999.
168
References____________________________________________________________
[102] Chow, K., J. P. Rode, D. H. Seib and J. D. Blackwell. Hybrid Infrared Focal
Plane Arrays, IEEE Trans. Electron Devices, Vol.29, No.1, pp3-13. 1982.
[103] Doudoumopoulos, N. A., D. J. Gulbransen, J. K. Kojiro, A. E. Cosand, C. G.
Whitney, R. Wyles, D. Murphy and S. Black. Infrared Readout Integrated
Circuit Technology for Tactical Missile Systems, In Proc. Meeting In Infrared
Readout Electronics, Apr. 21-22, 1992, Orlando, USA, pp.201-211.
[104] Wada, H., M. Nagshima, N. Oda, T. Sasaki, A. Kawahara, M. Kanzaki, Y.
Tsuruta, T. Mori, S. Matsumoto, T. Shima, M. Hijikawa, N. Tsukamoto and H.
Gotoh. Design and Performance of 256×256 Bolometer-type Uncooled
Infrared Detector, Proc. SPIE, Vol.3379, pp.90-100. 1998.
[105] Brady, J., T. Schimert, D. Ratcliff, R. Gooch, B. Ritchey, P. McCardel, K.
Rachels, S. Ropson, M. Wand, M. Weinstein and J. Wynn. Advances in
Amorphous Silicon Uncooled IR Systems, Proc. SPIE, Vol.3698, pp.161-167.
1999.
[106] Wu, C. Y. and C. C. Hsieh. New Design Techniques for a Complementary
Metal-Oxide-Semiconductor Current Readout Integrated-Circuit for Infrared
Detector Arrays, Optical Engineering, Vol.34(1), pp.160-168. Jan. 1995.
[107] Hsieh, C.-C., C.-Y. Wu, T.-P. Sun, F.-W. Jih and Y.-T. Cherng. High-
Performance CMOS Buffered Gate Modulation Input (BGMI) Readout
Circuits for IR FPA, IEEE J. Solid-State Circuits, Vol.33, No.8, pp.1188-1198.
1998.
[108] Chin, S. Y. and C. Y. Wu. A 10-bit 125-MHz CMOS Digital-to-analog
Converter with Threshold-voltage Compensated Current Sources, IEEE J.
Solid-State Circuits, Vol. 29, pp.1374–1380. Nov. 1994.
169
References____________________________________________________________
[109] Ringh, U., C. Jansson, C. Svensson and K. Liddiard. CMOS Analog to Digital
Conversion for Uncooled Bolometer Infrared Detector Arrays, Proc. SPIE,
Vol.2474, pp.88-97. May, 1995.
[110] Ringh, U., C. Jansson, C. Svensson and K. Liddiard. CMOS RC-oscillator
Technology for Digital Readout from an IR Bolometer Array, In Proc. of the
8th International Conference on Solid-state Sensors and Actuators, and
EuroSensors IX, June 25-29, 1995, Sweden, pp.138-141.
[111] Xie, G. Z., L. Deng, T. Wang, J. H. Xu, X. D. Chen, Y. D. Jiang, Z. M. Wu, C.
H. Huang and B. C. Yang. Readout Circuit Design for Uncooled Infrared
Focal Plane Arrays, In Proc. of IEEE 11th International Symposium on
Electrets, Oct. 2002, Australia, pp.349-352.
[112] Marshall, C. A., N. R. Butler, R. Blackwell, R. Murphy and T. Breen.
Uncooled Infrared Sensors with Digital Focal Plane Array, Proc. SPIE, Vol.
2746, pp.23-31, 1996.
[113] Altman, M. P., B. Backer, M. Kohin, R. Blackwell, N. R. Butler and J. H.
Cullen. Lockheed Martin's 640×480 Uncooled Microbolometer Camera, Proc.
SPIE, Vol.3698, pp.137-143. 1999.
[114] Mandl, W. J. Advances in On-focal-plane A/D with Low-power Optical
Readout, Proc. SPIE, Vol. 3379, pp.473-482. 1998.
[115] Mandl, W. J., J. J. Kennedy and M. Chu. MOSAD IR Focal Plane per Pixel
A/D Development, Proc. SPIE, Vol. 2745, pp.90-98. 1996.
[116] Ringh, U., C. Jansson and K. C. Liddiard. Readout Concept Employing a
Novel On-chip 16-bit ADC for Smart IR Focal Plane Arrays, Proc. SPIE, Vol.
2745, pp.99-110. 1996.
170
References____________________________________________________________
[117] Murphy, D., W. Radford, J. Finch, A. Kennedy, J. Wyles, M. Ray, G. Polchin,
N. Hua and C. Peterson. Multi-spectral Uncooled Microbolometer Sensor for
Mars 2001 Orbiter THEMIS Instrument, In Proc. 2000 IEEE Aerospace
Conference in Big Sky, 2000, Montana, pp.151-163.
[118] Parrish, W. J. and J. T. Woolaway. Methods and Circuirty for Correcting
Temperature-induced Errors in Microbolometer Focal Plane Array, U.S. Patent,
No. 5756999, 1998.
[119] Parrish, W. J. and J. T. Woolaway. Methods and Circuirty for Correcting
Temperature-induced Errors in Microbolometer Focal Plane Array, U.S. Patent,
No. 6028309, 2000.
[120] Howard, P. E., J. E. Clarke, A. C. Ionescu, C. C. Li and J. C. Stevens. DRS
U6000 640×480 VOx Uncooled IR Focal Plane, Proc. SPIE, Vol. 4721, pp.48-
55. 2002.
[121] Parrish, W. J. and J. T. Woolaway. Improvements in Uncooled Systems Using
Bias Equalization, Proc. SPIE, Vol. 3698, pp.748-755. 1999.
[122] Thodberg, H. H. Review of Bayesian Neural Networks with an Application to
Near-infrared Spectroscopy, IEEE Trans. Neural Networks, Vol.7, No.1,
pp.56–72. 1996.
[123] Panicacci, R., S. Kemeny, L. Matthies, B. Pain and E. R. Fossum.
Programmable Multiresolution CMOS Active Pixel Sensor, Proc. SPIE, Vol.
2654, pp.72–79. Feb. 1996.
[124] Cannata, R. Introduction to Infrared Focal Plane Electronics, The SPIE Short
Course Note, San Diego, CA, July 1993.
171
References____________________________________________________________
[125] Chiu, C. F. and C. Y. Wu. The Design of Rotation-invariant Pattern
Recognition Using the Silicon Retina, IEEE J. Solid-State Circuits, Vol.32,
pp.526–534. Apr. 1997.
[126] Baxter, C. R. and M. A. Massie. On-chip Signal Processing Configurations for
Focal Plane Arrays, Proc. SPIE, Vol. 3698, pp.726-735. 1999.
[127] Mizaikoff, B. Recent Trends in Mid-infrared Sensing, Proc. SPIE, Vol.4253,
pp.76-83. 2001.
[128] Tanaka, A., S. Matsumoto, N. Tsukamoto, S. Itoh, K. Chiba, T. Endoh, A.
Nakazato, K. Okuyama, Y. Kumazawa, M. Hijikawa, H. Gotoh, T. Tanaka and
N. Teranishi. Influence of Bias-heating on a Titanium Bolometer Infrared
Sensor, Proc. SPIE, Vol. 3061, pp.198-209. 1997.
[129] Hegel, Jr. R. R., R. A. Wood. Array Uniformity Correction, U.S. Patent, No.
4752694, 1988.
[130] Kaisha, T. K. Non-contacting Thermometer, European Patent, No. EP
0798545A1. 1997.
[131] Gu, X., G. Karunasiri, J. Yu, G. Chen, U. Sridhar and W. J. Zeng. On-chip
Compensation of Self-heating Effects in Microbolometer Infrared Detector
Arrays, Sensors & Actuators A, Vol. 69, pp. 92-96. 1998.
[132] Ramakrishna, M. V. S., G. Karunasiri, P. Neuzil, U. Sridhar and W.J. Zeng.
Highly Sensitive Infrared Temperature Sensor Using Self-heating
Compensated Microbolometers, Sensors & Actuators: A, Vol. 79, pp.122-127.
2000.
172
References____________________________________________________________
[133] Jansson, C., U. Ringh and K. Liddiard. Theoretical Analysis of Pulse Bias
Heating of Resistance Bolometer Infrared Detectors and Effectiveness of Bias
Compensation, Proc. SPIE, Vol. 2552, pp.644-652. 1997.
[134] Auerbach, F. J., G. Meiendres, R. Muller and G. J. E. Scheller. Simulation of
the Thermal Behaviour of Thermal Flow Sensors by Equivalent Electrical
Circuits, Sensors & Actuators A, Vol.41-42, pp.275-278. 1994.
[135] Jakonis, D., C. Svensson and C. Jansson. Readout Architectures for Uncooled
IR Detector Arrays, Sensors & Actuators A, Vol.84, pp.220-229. 2000.
[136] Xu, Y. P., X. B. Qian and G. Karunasiri. Circuit for microbolometer bias-
heating cancellation, Electronics Letters, Vol. 36, pp.1993-1994, Nov. 2000.
[137] Gregorian, R. and G. C. Temes. Analog MOS Integrated Circuits for Signal
Processing, Chapter 7, John Wiley and Sons, 1986.
[138] Qian, X., Y. P. Xu and G. Karunasiri. Self-heating cancellation circuits for
microbolometer, Sensors & Actuators A, Vol.111, pp. 196-202. 2004.
[139] Razavi, B. Design of Analog CMOS Integrated Circuits, pp.377-404,
McGraw-Hill Higher Education, 2001.
[140] Zeng, W.-J. Infrared Sensors Using Silicon Microstructures, Thesis of Master
Engineering, National University of Singapore, 1998.
[141] National Instruments. DAQ 6023E/6024E/6025E User Manual, Dec. 2002.
[142] National Instruments. LabVIEW Function and VI Reference Manual, Jan. 1998.
173
Appendix A.___________________________________________List of Publications
List of Publications
[1] Xinbo Qian, Yong Ping Xu and Gamani Karunasiri. Self-heating cancellation
circuits for microbolometer, Sensors & Actuators A, Vol.111, pp. 196-202.
2004.
[2] Xu, Y. P., X. B. Qian and G. Karunasiri. Circuit for microbolometer bias-
heating cancellation, Electronics Letters, Vol. 36, pp.1993-1994, Nov. 2000.
[3] X-B. Qian., Y. P. Xu and G. Karunasiri. A Prototype ROIC with bias-heating
cancellation for microbolometer IR FPA, Technical Digest of 27th international
conference on IRMMW, Sept. 2002, San Diego, USA, pp. 259-260.
[4] X-B. Qian, Y. P. Xu and G. Karunasiri. A tunable bias heating cancellation
circuit for microbolometer readout electronics, Proc. Sensor for Industry
Conference (SIcon’01), Nov. 2001, Chicago, USA, pp.264-268.
[5] Yong Ping Xu and Xinbo Qian. Current-mode Self-heating Cancellation
Circuits for Microbolometers, U.S. Provisional Patent, No. 60498465, Aug.
2003.
174
Appendix B.___________________________________________LabVIEW Programs
Appendix B. LabVIEW Programs
B. 1. Descriptions of sub-VIs used in imaging acquisition programs
: Configures an analog input operation for a specified set of channels.
: Sets the channel and scan clock rates.
: Starts a buffered analog input operation.
: Sets the output logic state of a digital line to high or low on a digital channel.
: Creates an n-dimensional array in which every element is initialized.
: Reads one frame data in the buffer and send the data out as 1-D array.
: Constructs the 2-D image array from the input 1-D data.
: Calculates the array size.
: Changes the dimension of an array, such as changes a 2-D array to 1-D array.
: Displays a dialog box that contains a message and a single button.
: Writes data string of a 2D or 1D array to a spreadsheet file.
: Reads a specified number of lines or rows from a spreadsheet file.
: Clears the analog input task.
: Clears the digital I/O task.
: Generates a color table to set the color of the front panel object.
: Computes the mean, standard deviation, and variance of input data sequence.
: Searches for the first maximum and minimum values in numeric array.
: Finds the discrete histogram of the input data sequence.
175
Appendix B.___________________________________________LabVIEW Programs
B.2. LabIVEW program for calibration operation
Calibration operation program (Frame 0)
Calibration operation program (Frame 1)
176
Appendix B.___________________________________________LabVIEW Programs
Calibration operation program (Frame 2)
Calibration operation program (Frame 3)
177
Appendix B.___________________________________________LabVIEW Programs
Calibration operation program (Frame 4)
Calibration operation program (Frame 5)
178
Appendix B.___________________________________________LabVIEW Programs
B.3. LabIVEW program for real-time imaging operation (Frame 0)
Real-time imaging operation program (Frame 1)
Real-time imaging operation program (Frame 1)
179
Appendix B.___________________________________________LabVIEW Programs
Real-time imaging operation program (Frame 2)
Real-time imaging operation program (Frame 3)
180
Appendix B.___________________________________________LabVIEW Programs
Real-time imaging operation program (Frame 4)
Real-time imaging operation program (Frame 5)
181
Appendix B.___________________________________________LabVIEW Programs
Real-time imaging operation program (Frame 6)
.4. LabIVEW program for noise measurements
B
182